You are on page 1of 372

The History of Mathematical Tables

FROM SUMER TO SPREADSHEETS

The History of
Mathematical Tables
FROM SUMER TO SPREADSHEETS

Edited by
M. Campbell-Kelly
M. Croarken
R. Flood
and
E. Robson

OXFORD
UNIVERSITY PRESS

This book has been printed digitally and produced in a standard specification
in order to ensure its cantmuing availability

OXFORD
UNIVERSITY PRESS
Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford
It furthers the University's objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lurnpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan South Korea Poland Portugal
Singapore Switzerland Thailand Turkey Ukraine Vietnam

Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
© Oxford University Press
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
Reprinted 2007
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
And you must impose this same condition on any acquirer

ISBN 978-0-19-850841-0

In memory of

JOHN FAUVEL
1947-2001

This page intentionally left blank

History of actuarial tables 79 CHRISTOPHER LEWIN AND MARGARET DE VALOIS 4. The computation factory: de Prony's project for making 105 tables in the 1790s IVORGRATTAN-GUINNESS 5. Difference engines: from Müller to Comrie 123 MICHAEL R. and 19 Assyria. Table making in astronomy 177 ARTHUR L. WILLIAMS 6. The General Register Office and the tabulation 209 of data. The 'unerring certainty of mechanical agency': machines 145 and table making in the nineteenth century DORON SWADE 7. Babylonia. Contents Introduction 1 1. 2500 BCE-50 CE ELEANOR ROBSON 2. The making of logarithm tables 49 GRAHAM JAGGER 3. 1837-1939 EDWARD HIGGS 9. Table making by committee: British table 235 makers 1871-1965 MARY CROARKEN . NORBERG 8. Tables and tabular formatting in Sumer.

Viii CONTENTS 10. The making of astronomical tables in HM 295 Nautical Almanac Office GEORGE. A. The rise and rise of the spreadsheet 323 MARTIN CAMPBELL-KELLY Biographical notes 349 Index 353 . Table making for the relief of labour 265 DAVID ALAN GRIER 11. WILKINS 12.

This page intentionally left blank .

.

For at least the last two millennia they have been the main calculation aid. Fig. Ready reckoners were pocket. to be brought forward from the narrow floodlights of particular special studies into the open sunlight. in the days before electronic computing devices. (iii) imperial Coinage (presenta- tion of empirical data). 0. and coinage. (iv) Money Lying at Undrawn Interest (theoretical data for use as a calculation aid). especially m multi-base Imperial systems of weights. and in dynamic form remain important today. . The history of tables now deserves. numerals. Introduction Tables have been with us for some 4500 years. Like other apparently simple technological or conceptual advances (such as writing. (ii) Weights of Materials (presentation of theoretical data). Their importance as a central component and generator of scientific advance over that period can be underestimated by sheer familiarity. or money) their influence on history is very deep. designed to aid everyday calculations. This page shows four different kinds of table: (i) The Metric System (empirical data for use as a calculation aid). published in London in the early twentieth century.1 A page from J. Gall Inglis's The rapid ready reckoner.or handbag-sized books of tables. measures. and is ready.

From the start. The history of this extraordinary human invention falls into four periods. dynamic tables in computers. Oriental studies. was the heyday of work on logarithm tables which formed the basis of calculation needs for the industrial revolution. and the histories of technology. and economic purposes.What they have in common is an expression of complex information in a two-dimensional form. physicists. has seen a number of developments in the produc- tion of a range of ever more sophisticated tables for physical. actuaries. The third period. printer. and consumer. The fourth period. It is now timely to seek to synthesize the results of these studies for a broader audi- ence and present for the first time a historic sweep ot a fascinating and unexpec- tedly important human invention. as well as the development of technology to help in their calculation. From the earliest times there has been a range of different kinds of table. mathematical. In the history of information processing. scribe. from the representation of mathematical func- tions to documents summarizing empirical values. A number of diverse roles are involved in the panorama of activities associated with tables. From around 2500 BC to 150 AD. whether scientific users such as astronomers. or navigators. work of high quality which has appeared in specialist journals in different areas (such as economic history. in the astronomical and trigonometric tables which lay at the heart of progress in the hard sciences leading up to the scientific revolu- tion. shows that there is still a lot of life in the deep idea of presenting information on a two-dimensional tabular screen. In recent decades there has been unprecedented historical activity on each of these four periods. The story is by no means over. and medical statisticians. constructor. for example. the transition from one-dimensional to two-dimensional layout in the location of information has a far greater significance than might naively be expected. . and science). or professional and trade users such as engineers.2 INTRODUCTION The issues turn out to be very interesting. among mem theoretician. from the mid-nineteenth century up to the present. indicates something of the challenge to historians to make sense of an extremely rich field of understanding. astronomy. mathematicians. as the development of spreadsheets. information science. the story is one of the invention of the table as a concept and its realization in a number of forms for different purposes. from the early seventeenth to the mid-nineteenth centuries. mathematics. computing. industrial. Over the next millennium and a half the second period saw some of the great achieve- ments of the human mind. The insights brought to bear by recent historiography enable us to position the history of tables as a fascinating confluence of different aspects of human endeav- our. we can see table mak- ing in the modern era as lying at the historical junction between the factory-based production of physical goods and the office-based processing of information. The structure of tables. The wide range of consumers of tables. issues of design and legibility jostle with issues of abstract information processing.

and extrac- tion of data—but there are as many different applications of mathematical tables as societies which used them. checking.They were particularly common:—almost essen- tial—in retail emporia that unit-priced goods.9.4. The ready reckoner was not just for calculating purchases in retail estab- lishments. John died unexpectedly in May 2001. the appropriate table gives the almost instant answer of £3 17s 7^d. Oxford in September 2001. and the production of tables. INTRODUCTION 3 These paragraphs were written by John Fauvel at the start of our plan- ning meetings for 'Sumer to Spreadsheets: The History of Mathematical Table Making*. and 11) and statistical data (Chapters 3 and 8). Imagine walking into a draper's at the turn of the twentieth century and buying 13j yards of cur- tain material at 5s 9d a yard. when currency decimalization (in Britain) and inexpensive electronic cash registers and calculators made them obsolete. They continued to be used up to about 1970. we have chosen to let him have the first word. These little books. The ready reckoner became a commodity item during the late Victorian period. 7. were small. They were a common calculating aid in the Western world. even in an age when schools taught commercial arithmetic. when low-cost publishing made them very affordable. but particularly in Britain because of the difficulty of computing sterling amounts and quantities in imperial weights and measures (both of which were non-decimal). from natural logarithms and other purely mathematical functions (Chapters 2. calculation. Types of tables Tables facilitate the selection. the communities of table makers and consumers. This was not an easy calculation. the summer meeting of the British Society for the History of Mathematics held at Kellogg College. of which this book is the final outcome. It was a compendium . though that was certainly its dominant use. the ready reckoner. and 10) to astronomical ephemerides (Chapters 1. However. ready reckoner. How can we begin to make sense of this enormous range? Take. often published with titles such as The pocket. and simply to expand on three of the themes he mentioned: the different types of table. inexpensive sets of commercial tables. Not wanting to dilute his voice amongst our own. The twelve chapters of this book cover a bewil- dering variety: from Sumerian tables of squares (Chapter 1) to late twentieth- century spreadsheets (Chapter 12). categorization. for example.

it should be possible to make a rough and ready 'table of tables'. Interest tables. censuses do not produce objective results but reflect the ideals and prejudices of the commissioning society. Wage tables. At one end of the empirical—theoretical spectrum there is census data. At the opposite extreme there are tables of mathematical functions. and whether they are primarily used to calculate other things or are an end in themselves. and dry measures. Stamp duty. . However carefully designed the process is. Table to calculate Easter day. as described by Edward Higgs in Chapter 8. Here are examples of some of the typical tables they contained: Perpetual calendar. cate- gorizing them as to whether they are computed from empirical or theoretical data. but they have to then categorize the relatively free-form responses they receive into discrete cate- gories for tabulation. Tables for commis- sions and discounts. Metric equivalents of imperial weights and measures. as embodied in the minds and work of the census officials. Sines. (for instance a railway timetable)? In short. we can consider what tables are used for: are they aids to further calculation (such as the dreaded four-figure tables of school mathematics before the pocket calculator made them obsolete) or do they pres- ent self-contained information in a final form. First. Not only do census designers have to choose the types of information they wish to elicit from.4 INTRODUCTION of tables that any citizen—from cleric to jobbing builder-—might find use- ful to have about their person. the population. we can think about what goes into tables: are they computed from empirical or theoretical data? Or to put it another way. Postal rates. Currency conversion tables. to what extent are the contents of the table determined by social convention and/or objective mathematical formulae? Second. For example. liquid. Money Lying at Undrawn Interest Used to present data Imperial Coinage Weights of Materials But categorization is not straightforward. such as those manufactured by de Prony and his team in late eighteenth- century Paris (see Ivor Grattan-Guiness's description in Chapter 4). Classification and taxonomy offers one way of making sense of the universe of tables. the tables from The rapid ready reckoner shown in the frontispiece to this Introduction might be classified like this: Computed from Computed from theoretical empirical data data Used as calculating aid The Metric System. Areal.

of course. as Arthur Norberg explains in Chapter 7. Here we see an interesting paradox: on the one hand computers have been the death of the printed table-as-calculating-aid but conversely computerized spreadsheets have given new and vigorous life to the still ubiquitous table-as-data-presentation format. In Chapter 3 Chris Lewin and Margaret de Valois explain how actuaries used tabulated mortality data to create and develop the concept of life assurance. number base. Communities of table makers and users Table making is a community-based activity. should we consider astronomical ephemerides. INTRODUCTION 5 logarithms. Logarithms. Communities of table makers and table users are as old and diverse as tables themselves. and planetary motion. for instance. we now see a form of table that can both show empirical or derived data and then manipulate that data and display the results all in one and the same 'table'. or number theoretical series—provided they are calculated accur- ately—are (and have been) reproducible across time and space regardless of the language. are the archetypical calculating aid. Ephemerides were calculating aids too. How. On the calculation-aid/data-presentation spectrum we see a similar range of possibilities. However. Along the spectrum in between are myriad tables which are constructed from a mixture of empirical and theoretical data. as Graham Jagger shows in Chapter 2. or personal belief system of the tabulator. which tabulate the predicted movement of the major heavenly bodies over the coming year(s)? They depend on a celestial coordinate system and on a mathematical model describing lunar. on the whole. Census tables and tables of vital statistics were. When we move on to look at dynamic tables in the form of spreadsheets (Chapter 12). enabling navigators to locate their position on the Earth's surface by the night sky. which makes it in some sense more objective than a census table. both of which are socially defined and open to constant revision and refinement. On the other hand astronomers could also use them like a railway timetable. three . the socially constructed and the value-free. meant to be read as final reports rather than as raw material for further calculations—although this was necessary too for instance to enable civil servants to plan a community's needs. But from another viewpoint an ephemeris is no more than the tabulation of a mathematical function. solar. looking up the times and locations of celestial events that they wished to observe.

and had a major influence on European astronomy in the following centuries. The Tbledan tables became highly popular throughout Europe until the fourteenth century. assembled between 1263 and 1272 by a group of 15 astronomers at the instigation of Alfonso X of Leon and Castile.1 He selected the most useful of the astronomical tables from the Almagest. or zij. which revolutionized astronomical thought.78Q—850). were further devel- oped in England. The Renaissance saw huge advances in both theoretical and observa- tional astronomy in which the invention of the telescope and the gradual adoption of the Copernican view of the solar system played a fundamental . of its predecessors.3 Another collaborative medieval table-making effort was the Alfonsine tables. Their longevity. mathematical. Persian. and Sanskrit and had a circulation larger even than the Almagest. wide distribution and influence among astronomers worldwide mean that Ptolemy's Handy tables can justi- fiably claim to be the first mass produced mathematical table. The Handy tables were originally copied by hand in manuscript form but editions in print form continued to be published in Western Europe for centuries. Claudius Ptolemy (<r. It had a long life: Adelard of Bath. The Alfonsine tables gained popularity in Paris in the 1320s. edited and republished them with instructions on their use as the Handy tables. translated it into Latin in the early twelfth century-—and it was still used in the Jewish Geniza in Cairo in the nineteenth century!2 In the second half of the eleventh century. was author of the 13-volume astronomical work Syntaxis or Almagest. of the royal House of Wisdom in Baghdad. While he is best known as the father of modern algebra and algo- rithms. an important group of astronomers gathered in Toledo in Muslim Andalusia (southern Spain) to produce the Totedan fables—some parts of which derived from the work of Arab astronomers such as al-Battani and al-Khwarizmi and other parts from Ptolemy. working in the famous library at Alexandria in Egypt. al-Khwarizmi's collection of astronomical tables and accompanying instructions. Arabic. for instance. is the earliest surviving astronomical treatise in Arabic.85—165 CE). They were translated from Greek into Latin. and tables. and actuarial communities. Another famous table maker who was the product of a scholarly institu- tion was al-Khwariznii (c. Astronomers have proved to be the most enduring community of math- ematical table makers and table users as generation after generation has built upon the work.6 INTRODUCTION particular groups feature across many of the chapters of this book: the astronomical.

planets. The international nature of the astronomical community allowed for astronomers across Europe to learn from one another and to build upon each others* work. editor of the Nautical almanac 1767—1811. Moon. The interplay between changing astronomical the- ory and the generation of new astronomical tables is well described by Norberg in Chapter 7. It is here that we begin to see the division between those who actually computed the tables and the editors in whose name the tables were published. For example Nevil Maskelyne. another important. While astronomers have been the predominant makers and users of mathematical tables over the last four millennia. Ephemerides. not only checked the Nautical almanac tables against those of the French Connaisssance des temps but also kept up a vigorous astronomical correspond- ence with his French counterpart Jerome Lalande throughout the intermit- tent Anglo-French wars of the period. Gunter.4 By the 1930s national ephemeris publishers throughout Europe had divided up the necessary table-making work between the various international offices in order to reduce the work- load and avoid duplication of effort. and others. and stars. group is the mathematical community. often by a team of computers who were sometimes part of the astronom- ical research community and sometimes simply hired hands. Even today international cooperation is common and. Jagger's narrative demon- strates how. the mathematical community worked to develop the use of logarithms as a calculating aid and to prepare the necessary tables. International col- laboration in the construction of national ephemerides was common. and others led to increasingly accurate tables for predicting the positions of the Sun. Newton. They were constructed from the best available astronomical tables. INTRODUCTION 7 part. were highly collaborative works. and often linked. The work of Brahe. Mayer. In Chapter 2 Jagger describes Napier's invention of logarithms and the subsequent publication of logarithm tables by Briggs. as George Wilkins explains in Chapter 11. Kepler. the British and US almanacs are now jointly computed and published. Mathematical tables of the higher functions began to appear at the beginning of the eighteenth century and the increasing importance of higher mathematical functions to mathematical physics throughout the nineteenth century led to a rapid increase in the number of tables pub- lished. Mathematicians began to be concerned by the lack of coordination and unwitting duplication of table-making effort within the mathematical . following Napier's initial work. or tables of predicted positions.

is discussed by Lewin and de Valois in Chapter 3. At its inception the Mathematical Tables Committee was highly representative of the communities it served. Its founding members included Francis Baily and his brother Arthur. As described by Mary Croarken in Chapter 9. the project's leader. However. and Charles Babbage. cur- rency exchange expert Patrick Kelly. and astronomical communities occurred in early nineteenth-century Britain. In 1820 the Astronomical Society of London was founded with the aim of standardizing astronomical calculation and collecting and distributing data. Arnold Lowen. the mathematician and astronomer John Herschel. It is often assumed that actuarial table makers had little in common with either the mathematical or astronomical communities. The skills for producing actuarial tables were equally applicable to making astronomical or mathematical tables and vice versa. sought to gain approbation for the project within the American mathematical community.8 INTRODUCTION and mathematical physicist communities. Astronomer Royal 1720—1742. the financier Henry Thomas Colebrooke. Mathematical Tables Committee became a central focus for mathematical table making in Britain.5 Perhaps the most significant contact between the actuarial. this assumption proves to be much too simplistic. and touched on by Higgs in Chapter 8. In response the British Association for the Advancement of Science set up a Mathematical Tables Committee in 1871 to produce a survey of existing mathematical tables and to coordinate the creation of new ones. A similar separation can be found in the Works Project Administration Mathematical Table Project created in 1930s New York and described by David Grier in Chapter 10. albeit with limited success. mathematical. the mathematician son . and socially separate from the community of users they sought to serve. later on in its history we see a growing gulf between the mathematical table maker and the user commun- ities. For example Edmund Halley. culturally. was author of a highly influential life table published in 1693. their colleague Benjamin Gompertz. In several instances Mathematical Committee members were making tables primarily because they enjoyed the process of making tables and not wholly for the benefit of the target users. Here the table makers were not willing volun- teers but unemployed workers desperately in need of work. over the next ninety years the British Association. Another table-making community. the actuaries. successful London stock brokers. Here the table makers were intellectually. their products being used in different spheres of human endeavour.West India merchant Stephen Groombridge.

During both World Wars mathematicians and astronomers came together and applied their table-making skills to the preparation of ballistics and other tables required by the armed services. before one word was spoken. One can discern five distinct styles of table making: the solitary table maker. In Britain the staff of the Nautical Almanac Office were joined by gradu- ate mathematicians to create the Admiralty Computing Service which pro- duced almost one hundred tables of direct relevance to the war effort. Although this is the order in which these five different styles evolved. and. INTRODUCTION 9 of a London banker and briefly actuary of the Protector Assurance Society. though the tradition is perhaps exaggerated. uni- versities. Smith. Making tables A theme that surfaces in many of the chapters of this book concerns the way in which tables were actually calculated. it was said that when Napier and Briggs first met—the latter having completed the tables the former had begun-— 'almost one quarter of an hour was spent. Table making as a communal enterprise has thus been part of the history of mathematical tables from ancient times. mechanized computing. manual and mechanized computing styles have co-existed throughout the history of table making.'6 However. is remarkable. The predominance of businessmen. had practical experi- ence of actuarial table making as well as an interest in astronomy and math- ematics. but in the Second World War the net was stretched wider. but both modes clearly existed. During the First World War it was primarily mathematicians such as Karl Pearson in Britain and Oswald Veblen in the US. For example. solit- ary and communal. But the lone . most of whom. Today so many businesses. finally. computerization. arid individuals run Microsoft Excel spreadsheet software that the community of table users with the ability to share spreadsheet data on a world wide basis is truly worldwide. We do not know the extent to which tables were solitary or communal endeavours. Adversity too has shaped mathematical table-making communities. The solitary table maker is a heroic figure. E. The working methods of the early table makers are shrouded in mystery. each beholding the other with admiration. Napier at least had a little help from his friends. computing bureaus. communal computing. as Jagger writes in Chapter 2. according to D.

For exam- ple. by division of labour.10 INTRODUCTION table maker continued to work long into the twentieth century. a community of specialized workers was vastly more productive. much more usually the historian is left with just a tantalizing hint. 'an extremely cultured lady but by no means a mathe- matician'. table making in Mesopotamia was very much a collective activity. was occupied for several years in the solitary task of compiling and publishing her 500 page volume of Natural sines. Leibniz hints at the existence of major table-making proj- ects.8 Such clear evidence of the way computing is organized is rare. de Prony con- sciously adopted an industrial metaphor. In an evocative passage. than one solitary worker who undertook all the processes of pin manufacture. as explained in Chapter 1 by Eleanor Robson. from the earliest times big table-making projects had to employ several human computers. de Prony decided 'to manufacture logarithms as one manufactures pins. noting in 1745 that his newly invented calculating machine 'has not been made for those who sell vegetables or little fish. For example. De Prony"s innovation was path-breaking for two reasons. He had been influenced by Adam Smith's classic Wealth of nations (1759). but the surviving artefacts are necessarily silent about how the work was actually organized. Second. and their work was entirely coordinated though the postal system.' The Bureau de Cadastre . located all over England. In 1790. published continuously from 1767.7 However. the sheer scale of the activity: the new French tables of logarithms and trigonometrical functions were of an unprecedented preci- sion and were computed de now (unlike earlier tables which had generally been compiled from previously existing tables). per person. One of the best known examples of such a group endeavour is the Nautical almanac. many of whom served for decades. highly organized bureaucracy. in which he gave his famous descrip- tion of a pin-making factory where. the Astronomer Royal Nevil Maskelyne employed a network of computers. as Croarken explains in Chapter 9. Gaspard Riche de Prony established the Bureau du Cadastre to produce a new set of tables for the French ordnance survey. As described by Wilkins in Chapter 11."* In the twentieth century. the British Association Mathematical Tables Committee made use of small calculating teams as well as solitary calculation. but for observatories or halls of computers. The Opus Palatinum of Rheticus published in 1596 (and the source book for Gifford's Natural sines) was said to be the work of eight computers. informal computing teams to a large-scale. quoted by Grattan-Gtiinness in Chapter 4. Emma Gifford. The Bureau marked a transition from small. First. The computers were all freelance.

For example. which required only the operations of addition and subtraction. By employing the method of differences. computing labour. INTRODUCTION 11 was in every sense a computing'factory'. the Nautical Almanac Office's network of freelance computers was replaced by a permanent office of about a dozen computing staff. At its peak the project employed 200 computers. an executive group of between six and eight professional mathematicians. For reasons discussed by . and it came into its own with the computation of LORAN tables in the early years of the war. described by Grier in Chapter 10.This tide of table making. as explained by Higgs in Chapter 8.The Bureau employed three classes of labour: a small advisory group of eminent analysts. Census tabulation used the 'ticking method* which was extremely tedious but required only the lowest form of clerical life. The WPA project was established partly as a make-work project for the unemployed. for example. He was one of the foremost economists of his day. As Mike Williams explains in Chapter 5. which threatened to overwhelm existing methods. In the 1830s. the difference engine was intended to entirely remove the human element.1" The difference engine was the most radical departure from manual methods of table making. and thoroughly con- versant with the principle of the division of labour. In the case of the British Census. Charles Babbage was the most important figure in the history of the difference engine. The WPA project was by no means the only such computing organization in wartime America. and once set up an engine would churn out results indefinitely. led the Moore School to become the birthplace of the modern computer. but also as a genuinely needed computing service. Although table makers sometimes used calculating aids (such as logarithm tables or desk calculating machines). Organized computing bureaus became the norm for large-scale table-making projects. The computing bureau style of operation remained popular for major table-making projects until after the Second World War. in the 1820s human kbour was starting to be challenged by machinery and the pin-making factory was giving way to the pin-making machine. the Army's Ballistics Research Laboratory and Moore School of Electrical Engineering at the University of Pennsylvania employed a team of 100—200 female computers (each equipped with a desk calculating machine) to make ballistics tables. and a large number of between 60 and 80 relatively unskilled computers." However. it was possible to use the lowest grade. and therefore the most economical.The processing of cen- sus data was another example of the large-scale employment of low grade staff for table production. The best documented was the WPA Mathematical Tables Project established in New York in the 1930s depression.

This was clearly understood in the late 1920s by L.J. computing. the British Association Mathematical Tables Committee. The difference engine concept was never really com- mercially viable. It seems that table making was too specialized and too narrow an activity to support a dedicated technology. Computer power enabled an explosion in the production of tables—not all of them of obvious utility. Other difference engines followed. From Sumer to spreadsheets The history of tables is not just an arcane corner of the history of science (mathematics. By contrast. Babbage failed to complete a full- scale machine—being diverted by the analytical engine. by giving more individuals and institutions the capability to do com- puting. adapted commercial calculating machinery for the Nautical almanac. Herman Hollerith recognized that census data processing was not a big enough market for his punched card machines. However. George Grant in the United States. and Christel Hamann in Germany. by the time the British. Conine who. census came to use Hollerith machines in 1911.12 At first it was thought that com- puters. and so he developed equipment for ordinary businesses. notably Georg and Edvard Scheutz in Sweden. A copy of their machine was purchased by the General Register Office to make a new life table. came to terms with the fact that the life's work of its members was no longer needed. We have already touched upon . or astronomy).12 INTRODUCTION Williams. the need for tables—whether on punched cards or in printed form-—evaporated. by Martin Wiberg in Sweden. census tabulation was only one of many applications for punched-card machines. and then undermined their very existence. after the 1890 US Census. as described by Williams in Chapter 5. In Chapter 9 Croarken describes rather poignantly how one institution. many of which were barely used. would encourage the production of many new tables. and earned the computer the nickname 'Bessie*. and did only a fraction of the job. Thus. However. ushering in an era of mechanized table making at last. however. The emergence of the digital computer first transformed the way tables were made. For example the Harvard Mark 1 computer produced some fifty vol- umes of Bessel function tables. and special-purpose machines had died out by the First World War. and Doron Swade in Chapter 6. although the machine required much coaxing. when it became clear that computers could calculate function values on-the-fly. he inspired several imitators.

INTRODUCTION 13 some aspects of economic history in the way that labour and machinery were organized to construct tables.13 the table as a pre-modern phenomenon of structured thought has been completely neglected. While the list has been hailed as a major breakthrough in cognitive his- tory. and constructive use for millennia.14 For Goody. human thought.. Indeed.. 'with links not only to the Greek theoretical astronomy of Hipparchus and Ptolemy but also backward to the Babylonians and . Robson's chapter on the uses of numerical tables in many aspects of lit- erate life in ancient Iraq (Chapter 1) shows that tables are by no means an invention of modernity but have been in lively. But we can also examine the internal structure of tables. They tell us much about how people have selected.'5 and a similar number of astronomical tables have been recovered from the first five centuries CE.. written in Demotic Egyptian. and manuscripts to printed books and shrink-wrapped software packaging—sheds new light on the history of the book and the transmission of knowledge. and circulated. and Greek. or at least their producers and users do.. then. for Goody (non-numerical) tables are no more than a means for modern scholarship to present 'the communicative acts of other cultures. and what should be collected and presented. non-literate and literate. Higgs's chapter on the General Register Office's tabulation of data in nineteenth-century Britain does indeed show that'the construction of tables involves decisions about what is important and what is not.' whose 'fixed two-dimensional character of may well simplify the reality of oral commun- ication beyond reasonable recognition. The communities they worked in. simplifying order on the com- plexities of other cultures. It is not difficult to assemble examples from other ancient cultures. The material culture of tables-—from clay tablets. and manipulated quantitative data at different times and places. We have seen that tables have a social history too. Coptic. a way of organising knowledge about classificatory schemes. and hence decrease rather than increase understanding*. and what can be ignored* but that 'there is nothing necessarily sinister about such processes of truncation'. Over a hundred arithmetical tables are known from the world of Classical Antiquity. papyri. Tables also speak to us about the history of literacy and numeracy. played a major part in how tables were developed. as greatly overlooked evidence for cognitive history. and the support those communities gave or denied them. classified. most famously by the anthropologist of literacy Jack Goody. inventive. tables are a means by which Western academia imposes inappropriate. symbolic systems. published.

tables could be used to store basic functions used for more complex calculations.14 INTRODUCTION forward to the astronomers of Byzantium. Columns could be totalled. Islam. Column headings and row labels. The scribe no longer had to waste time and effort in describing each item accounted for. unchanged through the millennia. Right to the present day. of the basic theme of homogeneous values arranged in rows and columns. In Chapter 1 Robson traces the functional development of administrative tables in early Mesopotamia over a period of just 50 or 60 years in the nine- teenth and eighteenth centuries BCE. whether literary epics or royal decrees. and intervals reconsidered according to the dictates of . one could use the very structure of tables as an aid to calculation. For example. headings. First. The two-dimensional structure of the table was elaborated remarkably quickly. residual errors were gradually eradicated. while the supervisor could take in at a glance the finished document and the pattern of information it contained. Second. User surveys tell us that spreadsheets are most commonly used to write lists and to keep simple accounts: two basic human needs. even though it was to be millennia before titles. Categorization and data selection come first: these first tables were no more than multi-column lists of dif- ferent named types of empirical data. paper for clay tablet. in the blink of an eye in historical terms.*16 This list could perhaps be extended indefinitely. producing a great efficiency both in recording and retrieving structured information. in logarithmic tables. were an integral part of tables right from their inception. When we substitute. Bessel functions for squares of the integers. and subheadings were applied to works of connected prose. Within very short order—a decade or so—the computational power of tables was discovered. This remains true even in the age of the spreadsheet. products could be found across rows. one can see that the similarities between tables ancient and modern are much greater than their differences. Thus quantitative and qualitative information could be separated. the table can be seen as an elaboration. but could simply tabulate it in the correct row and column. improved and perfected by each succeeding generation. values trun- cated or extended. from integers and their squares or cubes to reciprocals. but it took millennia for the amazing power of the deceptively simple table to reach its zenith. or EU farming subsidies for tables of goats and sheep. Tables are quintessential cultural objects of the civilizations that created them. and the Latin West. say. cross-checking for errors became easier.

North. Haywood. Ttte Fontana history of astronomy and cosmology. 1. Notes 1. pp. an ancestral line could be drawn from the logarithms of Briggs to the four-figure tables used in schools in the early twentieth century. p. A source book in mathematics. INTRODUCTION 15 the day. We know that dif- ferent civilizations independently invented (or perhaps. loonier. Indeed. Princeton University Press. 8. Giflbrd. G. 47. Cambridge. E. vol..'Tabulating the heavens: computing the Nautical Almanac in eigh- teenth century Britain'. Chapin. 186—206 and idem. 184-5. Fontana. New York. as Swade describes in Chapter 6. 1970. An index of mathematical tables. 1994. J. London. D. Oxford. J.. 165-80 and D. L. E. Nevil M. Halley. New York. like the integers. 2nd edn. in The Dictionary of Scientific Biography. It is likely that. Cambridge University Press. Fletcher et aL. vol.'. D. 207-13. 1989. Notes and Records of the Royal Society 32 (1978). and each 'upgrade' com- bines the best of the old leavened with a little of the new. 4. M. 1962. at least in theory. 5. 2. 1959. As Martin Campbell-Kelly has noted in the closing chapter.'An estimate of the degrees of mortality of mankind. 7. discovered) the table. S. Manchester. Fontana. D. North. 565-610 and 654-6. Princeton. pp. Howse. 1994. pp.'Ptolemy'. Would the civilizations of another planet use tables? We like to think so. 1998. IEEE Annals of the History of Computing. 1914. Philosophical Transactions of the Royal Society 17 (1693).askelyne: the seaman's astronomer. The Fontana history of astronomy and cosmology. . 'Lalande and the longitude: a little known London voyage of 1763'. 6. such attention to typographical niceties made the printing part of Babbage's difference engine almost as complex as the calculating part. J. E. Smith. p. Dover Publications. Quote taken fiom A. Blackwell. the two- dimensional table is almost an historical necessity—suggested by the two dimensional writing surface common to all civilizations. Croarken. London. but by their user interface. forthcoming. 818. Ptolemy's Almagest. the use of rules and 'white space'. 3. Natural sines. 2. Tables have followed the fashions of the day with regard to typeface. This trend continues even into the modern spreadsheet—where competing brands are distinguished not by their function. and even the colour of paper.

1999.. 1999. 1815—1848. Princeton. Cambridge. London. 1977. pp. A. 1977. 1999. Cambridge. The domestication of the savage mind. 2nd edn. Cambridge. Mass. 1989. 10. Translation from The works of Charles Babbage. pp. 1980.16 INTRODUCTION 9. Cambridge. Campbell—Kelly). p. Howard Aiken: portrait of a computer pioneer. MIT Press. M. Oxford University Press. 4 A classification of astronomical tables on papyrus'. quote from p. J. 225—8.. The machinery question and the making of political economy. 16. 53—4. Cohen. 12. he refined the concept into a more nuanced variation known as the Babbage Principle. Cambridge University Press. 14. 11. Jones. Berg. which advocated that manufacturers should purchase no more than the lowest quality labour that could perform any given task. The domestication of the savage mind. Indeed. M. Cambridge. Zeitschriftjiir Papyrotogie und Epigraphik 105 (1995). in Ancient astronomy and celestial divination (ed. H. vol. 1972. 181. Pickering. 13. Goody. Goody. The computer. 229-340. 268—76 and 'Further arithmetical tables'. . J. B. See die catalogues by D. Cambridge University Press. H. Fowler in The mathematics of Plato's Academy: a new recon- struction. M. Oxford. Swerdlow). 2 (ed. 15.from Pascal to von Neumann. Mass. Princeton University Press. Goldstine. 335. N. MIT Press. Cambridge University Press. pp. H. I.

This page intentionally left blank .

.

2500 BCE-50 CE ELEANOR ROBSON Fig. recording the monthly salaries of forty-five temple personnel.1 ftonth-by-month wage account for the temple of Enlil at Nippur. There are also subtotals for each person every six months and a yearly total. There are explanatory interpolations under two headings and under their totals. 1906. The University Museum. pis. and Assyria. Babylonia. most of these are classified (in the penultimate column) as dead or absconded. Eighteen of the individuals listed get no payment for all or half the year. T. (CBS 3323. Documents from the temple archives of Nippur dated in the reigns of the Cassite rulers. A. exhibits most of the classic features of Mesopotamian tables. now in the University Museum. 1. and a summary at the end..) . Philadelphia. Names and professions are given in the final column. 3 vols. from Nippur. University of Pennsylvania. 25-6. for the year 1295 BCE. Column headings at the top of the table give the month names. 1 Tables and tabular formatting In Sumer. This tablet. Clay.

especially in mathematics and metrology. down which the data is attributed to different individuals or areas. contexts. disappearance. while unheaded tables do not. and the vertical axis. documents with tabular formatting could be found in three distinct Mesopotamian locales: in the large institutional administrative archives of Sumer and Babylonia. later. any qualitative or descriptive information is almost invariably contained in the final right-hand column (row labels). on the other hand. A S S Y R I A While most of this book Is concerned with the development and production of standardized mathematical tables for use as reference tools. and cultures. without explicit delimiters. if we are to talk about tables clearly and effectively we need a consistent terminology to describe them. While some tables have two axes of calculation others exhibit just one. partial adoption. Broadly speaking.20 TABLES I N SUMER. or interrupts the table as an explanatory interpolation. Both calculation and organization run. in the scholarly libraries attached to the great temples of Assyria and Babylonia. with clearly traceable consequences across an ever-widening arena of functions. But first. even within the single cultural milieu of ancient Mesopotamia we shall see a fitful pattern of invention. infor- mal tables. left to right and top to bottom. or none at all. Mesopotamian tables have at most two axes of organization: the horizontal axis. as a rule. In Mesopotamian tables. others are followed by summaries and/or colophons. The 5600 documents from late fourth-millennium Uruk and its neighbours *serv[ed] the accounting needs of a complex administration including offices . Some tables are preceded by titles or introductory preambles. Headed tables have columnar headings. We shall examine all three in turn. This was not a one-off event. separate quantitative and qualitative data by spatial arrangement only. Tabular formatting in administrative records There are no tables to be found in the earliest written records of Sumer. Documents with no tabu- lar formatting at all we might call prose-like or prosaic. this chapter will focus more on the invention and evolution of the numer- ical table as an information storage device. and re-invention time after time over the course of some two and a half millennia. and. B A B Y L O N I A . along which different types of numerical information are categorized. I shall define a formal table as having both vertical and horizontal rulings to separate categories of information. Rather. following the direction of cuneiform script. amongst the detritus of scribal schooling. axis of calcula- tion. usually vertical.

5 billy goats 7 nanny goats: day 13 4 kids.The spatial organization of the cases on the sur- face of the tablet could be quite complex. in both senses of the word: they record the day to day transfers of goods. TABLES IN S U M E R . The visual logic of the earliest accounts was thus irrevocably lost in the shift to linear organ- ization of writing during the first half of the third millennium BCE. and personnel from the responsibility of one official to another. Nevertheless. 11 billy goats. royal deeds. 5 kids.1 They were formatted not into lines of ordered text hut into boxes or'cases'. of field management. accounting continued to dominate the written record: a recent estimate has put the total administrative output at some 97 per cent of all surviving tablets from the third millennium. 7 nanny goats: day 206 . 1 nanny goat 1 billy goat: day 14 2 sheep.5 The accounts are almost exclusively prosaic. and of labor'. But once writing evolved the capacity to represent syllables and thereby approximate the sounds of speech.2 reflecting often sophisticated accounting procedures/ but nowhere do we see the separation of quantitative and qualitative data into separate cases. 5 kids. and even poetry—it became increasingly imperative to order the written sym- bols to follow the structure of Sumerian as a spoken language. 1 billy goat 6 nanny goats: day 18 12 sheep. of herded animals and animal products. live- stock. and those records are made linearly across the surface of the tablet. grain production and processing. 9 nanny goats: day 16 3 sheep. for instance. each of which represented one sense unit or accounting unit. which could be displayed in any order within each case. 15 kids 3 billy goats. This earliest writing comprised just number signs and ideograms. representing whole words or ideas. B A B Y L O N I A . it was no longer determined solely by the needs of accounting.4 The bureaucracy of the twenty-first century BCE kingdom of Ur was particularly prolific: some 45 500 administrative records have so far been published from its archives in the ancient cities of southern Iraq. 3 kids 2 nanny goats 4 billy goats: day 15 3 sheep. is an extract from a monthly summary of four different types of ovids under the responsibility of the central livestock depot at Puzrish-Dagan: 7 nanny goats 3 billy goats: day 10 5 sheep. As literacy spread into other social domains—for recording legal decisions. A S S Y R I A 21 of the fisheries. This.

Akkadian is a Semitic language. long-lived and charismatic kings. 2350-2250 BCE) and Ur (c. the Tigris and Euphrates. centred on one city but aiming to unify die whole land. and most sophisticated in the world. Syria. indirecdy related to Hebrew and Arabic. who seem to have constructed their empires through sheer force of personality. but by the middle of the fourth millennium BCE the cities of Mesopotamia were the largest. 2100-2000 BCE) were each created by die ambitions and abilities of individual. The shift towards urban living was a long and complex process. wealthiest. Inherently unstable. Ancient national. or coies. It encoded the Sumerian language—related to no other known lan- guage. and standardization followed by unsteady but inexorable decline over the next two or three generations. and cultural boundaries were very fluid. the tides drew their strength from local religious power and from long-distance trade of agricultural and animal products in exchange for metals and luxury goods. were arranged in boxes. centralization. corresponding more or less to the modern state of Iraq. as the Gulf receded and people moved south into the gradually more habitable marsh- lands it left behind. . For the first time. A second language also came iuto widespread use in the course of the third millennium. representing whole words. regional. and each centred around the temple of an anthropomorphic deity. and bears no resemblance at all to Sumerian. and Iran. Mesopotamia was* first settled some eight thousand years ago. B A B Y L O N I A . During the Early Dynastic period (e. This enabled the scribes of Sumer to record a wide variety of docu- ments and texts: they were no longer restricted to keeping accounts.22 T A B L E S I N SUMER. how- ever. these kingdoms both have histo- ries of rapid unification. which evolved the capacity to represent syllables and thereby approximate die sounds of speech. 3000—2400 BCE) the incised script developed into wedge- shaped or cuneiform characters. A S S Y R I A Early Mesopotamia: Sumer and Babylonia Mesopotamia is the classical Greek name for the Land Between Two Rivers. Sargon and Shulgi. At die same time. More or less politically autonomous. while retaining the ancient impressed notation for numerals. The last third of the third millennium BCE saw some major shifts in Mesopotamia!! socio-political structure. large territorial states came into being. and at times Mesopotamia incorporated neighbouring lands which now belong to modern Turkey. cases evolved into ordered lines of characters. by about 3300 BCE. The kingdoms of Akkad (c. had matured into die world's first writing. although this was still a core activity. Driven by-the need to manage complex economies and large populations of people and livestock. living or dead—with marks in the surface of hand-held clay tablets. the temple administrators of the city of Uruk developed an accounting system which. Originally impressed number signs and incised pictograuis or more abstract sym- bols.

reli- gious centre of Enlil. the 'father of the gods'. 1792-1750 BOB) vast swathes of southern Mesopotamia had come under Babylonian influence or control. or to see at a glance which days there are no entries for: given that this document would have been compiled from at least thirty daily records and might itself have been one of up to five hundred tablets from which an annual account had to be compiled. intelligence. changed hands between the south- ern kingdoms of Isin and Larsa at least half a dozen times in two centuries. By the end of his forty-year reign (c. B A B Y L O N I A . ASSYRIA 23 Early Mesopotamia: Sumer and Babylonia cont The fall of the kingdom ofUr at the end of the third millennium saw die political landscape of Mesopotamia revert to its earlier configuration of small city states vying for land. water rights. each at different stages of their education. Much of the learning was through repeated copying and rote memorization. whose primary profession was often that of a temple administrator or priest. Hammurabi also made sweeping religious changes. For the city of Nippur under Hammurabi's successor Samsu-iluna we can closely identify the curriculum used in three or four different scribal schools. none of Hammurabi's successors appears to have had his longevity. However. By the end of the sixteenth- century Babylon was little more than a city state again. They catered for up to half a dozen boys at any one time. housed in the home of the teacher. In the early eighteenth century Hammurabi. so that as before Mesopotamia was united only for a matter of decades before a series of contractions and retrenchments. and the variations between them.7 It too is an account of sheep and goats: 3 3 3 2 I lambs 93 93 93 6[2] 31 first-rate sheep 6 6 6 4 2 billy goats 102 102 102 68 P14 . this non-tabular format was gravely inefficient. The city of Nippur. often playing one ruler off against another. and trade routes. king of the hitherto unimportant city of Babylon. TABLES IN SUMER. one of which is also from Puzrish-Dagan. It is not an easy task to total each category of animal. A tiny number of tables are known from the bureaucratic archives of Ur. Scribal schools were for the most part very small indeed. managed to outwit and outmanoeuvre his rival dynasts in this real- life game of Monopoly. or stam- ina. culminating in the 'retirement' of Enlil in favour of his personal deity Marduk as head of the pantheon.

2050-2025 BCE. M. 1984. a purpose-built depot near Nippur. makes no visual separation between numerical and descriptive data. from Puzrtsh-Dagan. all of them excavated illicitly and therefore without context. Berrien Springs. two axes of organi- zation. Maybe the scribes were simply under too much pressure preventing institutional col- lapse—part of the general 'relatively sudden retreat of the central adminis- tration'9 at that time—to experiment further with new data storage formats. Almost 12000 documents from its archives have been published. no. Neo-Sumerian account texts in the Horn Archaeological Museum. The date is on the top of the reverse. . nor between different categories of data. one. c. On the right is the earliest known tabular account. no. Berrien Springs. ASSYRIA Fig.2 Compare these two accounts of sheep and goats. so that we cannot tell why the three varieties of livestock were further categorized. On the left the non-tabular account. Sigrist. Michigan. vertical axis of calculation.) In the fifth row the columns are labelled with the abbreviated names of the officials responsible. based on its probable archival context. It has most of the features of later administrative tables. Our table conies from the penultimate active year of Puzrish-Dagan.0400.0639 and ADAM 73. Summary information at the bottom of the tablet is largely missing.24 T A B L E S I N SUMER. 63. vol. row labels to the right of the table— but no headings. 1. the additions are correct-—so why didn't the idea catch on? We can put forward a tentative hypothesis. Line 5 appears to contain column labels in the form of abbreviated per- sonal names. columnar totals are in line 4.8 functioned as the central livestock management centre for the kingdom of Ur for over thirty years (c. after which date the number of surviving records declines dramatically. Puzrish-Dagan. Hundreds of tablets from each year of its operation have survived. Row labels giving the type of animal are on the far right. Andrews University Press. 2056-2020 BCE). 1. now in the Horn Archaeological Museum. (AUAM 73. at the bottom is the date on which it was drawn up. This appears to be a reasonably successful attempt at table making—its structure is clear. from an annual average of 220 in the preceding decade to single figures. 56. B A B Y L O N I A . c. Michigan. 2028 BCE.

At this point (c. He made no calculations. None has an introductory title or preamble but almost all have two axes of calculation and interlinear explanatory interpolations.'utu-Qour'. At least three different methods were used to record the offerings and redistributions. until after the conquest of Larsa by Babylon. and beer) made to Ninurta and other deities and divine objects which were then redistributed to temple personnel. and all are headed uniformly. Very few tables are attested after the mid-1720s for over 70 years until the last half of the seventeenth century BCE.10 The first phase in their develop- ment is attested by a group of around two hundred tabular lists from the temple of the god Ninurta in Nippur. reflecting the fact that they have no more than three data columns— whereas the earlier administrative tables regularly had eight or nine. B A B Y L O N I A .'fine flour*.IM 'its name*.12 the first of which was a period of experiment and evolution (c. These . 1855—1836} and one change of gov- ernment. horizontal or vertical. aligned with the names of the human beneficiaries. TABLES IN SUMER. usually simple columnar additions.'shortbread*. the most successful and enduring of which spanned at least twenty years ((. In this final column he listed the names of the divine recipients and in the others the quantities of goods they received. 1758 BCE) the tables begin to be preceded by general introductory matter. and the tablets are predominantly in por- trait format. with at most one axis of calculation.The tables from the immediately following decade (the mid-1790s to 80s) exhibit by contrast a relatively rigor- ous standardization." The ancient dates on the documents themselves span some eighty years and six changes of political allegiance between the kingdoms of Isin and Larsa between 1871 and 1795 BCE. in which the scribes sometimes experiment with the format and layout of the documents. A S S Y R I A 25 It was not until the mid-nineteenth century BCE that tables began to be used with any consistency or continuity. heading them'bread*. 1850—1795). Almost all are in portrait orientation. All are much simpler organ- izationally than their precursors. Several groups of related tables can be detected. with the final heading (over the row labels) reading MU.BI. flour. Almost all of them are in landscape format. There is then a twenty-five year gap in the documentary record. but leaving the final column blank.'beer'. The archive's scribe ruled the obverse of the tablet vertically into six columns. Over four hundred tablets record regu- lar food offerings (bread. and on the reverse he never tabulated the redistributions but simply listed them prosaically. Outside the Nippur offerings archive we can detect three phases in the development of administrative tables.

They exhibit all the featural complexity of the eighteenth-century tables—headings. ° An administrative archive of some 600 tablets. explanatory interpolations. From the late second millennium onwards. that is simply because there is no evidence for administrative tables there before the first millennium BCE. subtotals. Kalhu. a more satisfying explanation is at hand. the evidence for tables is very slim. For whereas the Assyrian .'14 Some 33 percent of the records in that archive are tabular. B A B Y L O N I A .BI. we can get a sense of the strength of the written tradition. We can only speculate how this extraordinary stability was main- tained. two axes of calculation. tides and preambles. In other words they refer to the handling and disposition of the taxes after they had been collected.IM 'its name*— some four hundred years after the last datable table from eighteenth-century Nippur. and Nineveh used tables simply as columnar lists. with columnar totals at the end of the table. present from absent. 1360—1225. Tempting though it is to attribute the lack of tables in the first millennium BCE to some ill-defined decline in bureaucratic ability. in order to separate numerical data of different types: wine from beer. whereas their predecessors are mostly from Larsa and the south. The evidence for first- millennium Babylonia is even scarcer. No surviving exemplars exhibit horizontal calculations or any operations more complex than simple addition. Our next evidence for tabular documentation conies once again from Nippur. and the payment of the salaries of the store- house officials as well as of the priests. Administrators in the successive capital cities of Ashur. as although drawings of all the documents in the archive were pub- lished nearly a century ago. and others in the temple service. Nevertheless. After the mid-seventeenth century BCE the Mesopotamian historical record falls silent for about three hundred years. totals. and in the face of massive documentation. A S S Y R I A late tables are almost all from Sippar and the region around Babylon. spanning the years c. and a hint of the huge gaps in all aspects of our knowledge of Mesopotamian histoty. comprises "records of the receipts of taxes or rents from outlying districts about Nippur. no analytical work has yet been done on them. and shows a remarkable similarity to what had gone before.26 TABLES I N SUMER. Even in the period of Assyrian world domination. Aramaic became increasingly common as a written language. So far we have had nothing to say about northern Iraq. some 4 per cent of the total.15 Most were headed. of commercial transactions conducted with this property. even the final column heading MU. Just eighteen of the 450 published administrative records from Nineveh are tables.

tradition and tamper-proof qualities of tablets ensured their continuing use for legal documents and religious litera- ture. followed by the Sumerian word sa ('equal* and/or 'opposite"). almost all of them from houses and buildings which burned down in a city-wide fire in about 2600 BCE. the new media.While the table is organized along two axes. Only six rows are extant or partially preserved on. It is highly likely. 2600 BCE. on the other hand. c. that the vast majority of tabular accounts from the first millennium BCE perished long ago with the boards on which they were written. A S S Y R I A 27 and Babylonian dialects of the Akkadian language were still recorded with the increasingly cumbersome and recondite cuneiform script on clay tablets. TABLES IN SUMER. could be drawn up. then. written freehand with ink on parchment or papyrus—neither of which organic media survives at all in the archaeological conditions of Iraq. the horizontal multiplications. there is just one axis of calculation. however.17 but sadly we have no detailed context for this tablet because its excavation number was lost or never recorded. several specimens of which have survived. corrected and added to over time. Around a thousand tablets were excavated from Shuruppag.6km to 360m in descending units of 360m. the reverse. Cuneiform tablets had to be inscribed while the clay was fresh.16 and for which there is ample documentation in the cuneiform record. The tablet is ruled into three columns on each side with ten rows on the front or obverse side. for . B A B Y L O N I A . were also written on waxed wooden or ivory writing boards. They continue the table in smaller units.18 While at first sight this table appears simply to list the areas of square fields. Prior to the inven- tion of the sexagesimal place value system in the twenty-first century BCE there was no concept of abstract numeration: numbers were not thought of as inde- pendent entities but as attributes of concrete objects—the length of a line. 3. like modern-day ledgers. while the final column gives their products in area measure. from 300 to 60m in 60m steps. and then perhaps (in the damaged and missing lower half) from 56 to 6 in in 6 m steps. The first two columns of the obverse list length measures from c. The durability. Aramaic used an alphabet of just 22 letters. All three languages. Formal and informal tables in school mathematics The first securely datable mathematical table in world history conies from the Sumerian city of Shuruppag. whereas writing boards and ink-based media were particularly suited to keeping accounts. namely. in fact it is akin to much more abstract multiplication tables.

Major new communication. and customs of their native predeces- sors. 1550-1150} had taken power. A S S Y R I A Later Mesopotamia: Babylonia and Assyria The disintegration of Babyloiiiae kingdom in die late seventeenth century BCE marked the end of the old order. transport. and interpreted anew. became luxury must-have commodities that stimulated long-distance trade and contact between elites all over the Middle East from Turkey to Iran to Egypt. while glass. remaining significant employers. Horses enabled people and ideas to travel faster than ever before. and consumers in the Babylonian economy. producers. While the Kassites ruled over southern Iraq. and military technologies were developing that would henceforth enable huge empires to grow and flourish over centuries. and later iron. the city of Ashur on the Tigris began to assert its independence. the north of the land belonged to the Syrian empire of Mitanni. though. ruling family had its origins somewhere outside Mesopotamia they quickly adopted the languages.28 TABLES I N SUMER. showing places mentioned in this chapter In Babylon the Kassite dynasty (t. and gradually built up . in which tradi- tional texts were revised. expanded. Although the. Wap of the Middle East. It was a time of literary and scholarly innovation. The old city temples retained their traditional wealth and status too. From the mid-fourteenth century onwards. B A B Y L O N I A . religion.

but in 539 BCE it became subsumed into the unprecedented^ enormous Persian empire.Assyrian attempts to usurp Babylon as the cultural centre of Mesopotamia were never entirely successful.The even lines give the areas that result from squaring the lengths. or the quantity of sheep in a flock. 900— 650 BCE) its capital moved up die Tigris. extracting heavy tribute from those lands that were not yet undo: its direct control.5m—6m). Greek-speaking Macedonians. first to Kalhu and then to Nineveh. followed.19 It too is in three columns. Once again. Nineveh fell to the Babylonians and Medes in 612 BCE. B A B Y L O N I A . The table was not the only means available for expressing numerical relationships. By the early first millennium Assyria held vast tracts of the Middle East in its power. but these function more like newspaper columns than the columns of a table. BCE to square a number meant. as expected. Esarhaddon. the temple of Marduk in Babylon was die last refuge of a small handful of traditionalist priests and scholars. we have no exact information about its original context. It had survived political and cultural conquest by Persians. A S S Y R I A 29 Later Mesopotamia. but the style of its handwriting suggests that it is perhaps a few generations younger . 1200 BCE was large enough to attempt a take-over of Kassite Babylonia. I to 12 cubits (c. instance. Therefore in the third millennium. By the end. Odd lines of the text give length meas- urements from. 0. TABLES IN S U M E R . and Ashurbanipal ruled the Middle East from enormous palaces built at the expense of their subject peo- ples. Cuneiform survived the coming of alphabetic Aramaic in the early first millennium. in large part because of its continued high status. to construct a square area from two equal lengths. some 600 years later. though. In the heyday of the Assyrian empire (t. where Sargon and his successors Sennacherib. a kingdom which by c. but Ashurbauipal managed to create a definitive library of traditional Mesopotaniiaii works in his palace at Nineveh. the guardians of a late flowering of the cuneiform culture that had blossomed for over three thousand years. Babylonia and Assyria cont. and the later Parthians. Babylon enjoyed a brief period of independence and prosperity under Nebuchadnezzar and his successors. decorated with detailed depictions of their conquests and filled with. Although this event marked the definitive loss of Mesopotamian political autonomy it was by no means the end of its cultural history: the last datable cuneiform tablet is an astronomical record from Babylon written in 75 CE. though: a linear version has survived from the city of Adab. by the word sa. booty from their annual campaigns. in part through the theft of tablets from Babylonia and the employment of captive Babylonian scribes to write for him. at some level.

2600 see. 3600-360 m) and the final column contains the square area of their product. 119.20 Oddly enough. 1993. England. (VAT 12593. and R. we know nothing else at all. B A B Y L O N I A . from Shuruppag. London and Chigago. now in the Vorderasiatisches Museum. from Shuruppag. 1. The first two columns contain identical lengths in descending order from 600 to 60 rods (c.30 T A B L E S I N SUMER. where one might expect to find tables galore. and probably finished at I rod (6m). H. Archaic bookkeeping. frustratingly. Berlin. While there are many thousands of published and unpublished . fig. K. 2550 BOB). Damerow. The sequence continues on the reverse.) than the tablet from Shuruppag (c. we find a strong disinclination towards the truly tabular format in later school arithmetic. P. University of Chicago Press. It is signed by someone called Narnmah—about whom. Nissen. c.3 The world's oldest datable mathematical table. ASSYRIA Fig.

Little attempt was made to align the numbers themselves. The series ran from tiny units to enormous ones. A S S Y R I A 31 documents from the early second millennium BCE that are normally char- acterized as mathematical tables. sometimes followed by the square and square root of the head number. Their informally tabular appearance derives from the fact that each line is left—right justified. more verbose forms of the list were copied on first exposure to a new section. The multiplicands were always 1—20. at the time of the first sexages- imal numeration. but never with column headings.26 The lists are usually informally tabular. as are the lines on all well written cuneiform. or lists of geometrical. like the multi- plication lists. while a terser—and often much more untidy—version. which are highly likely to have originated in the adminis- trative archives of the twenty-first century kingdom of Ur.25 and inverse cubes from 1 to 30. The whole sequence could comprise up to forty lists. or lists of inverses. The so-called coefficient lists. metrological. area. with head numbers from 50 down to 1.23 Less frequently extra-curricular arithmetical lists were copied. B A B Y L O N I A .15. Different multi- plications were omitted in different schools. As usual. on closer inspection most of them turn out to be prosaic lists of equivalences rather than formally laid out tables. such as reciprocal pairs halved and doubled from an initial pair 2 05 ~ 28 48. At around the same point in the curriculum. 30. from the capacity. tablets. was repeated for revision. In general. TABLES IN SUMER. through repeated copying and revision. The same type of reciprocal list is very well attested in later centuries.22 They list sexa- gesimally regular numbers up to 60. The students were required to memorize it. are similarly . although occasionally formal columnar layouts are adopted. and 50. with their inverses. The earliest exemplars of the second-millennium arithmetical tradition are actually from the tail end of the third millennium BCE. 40. each of up to 25 lines long—some 1000 lines in total. but the sequence was universally the same. or sometimes beyond. like all other ele- mentary school subjects. the evidence is patchy and difficult to date accurately. it appears. and at its maximum extent had about 600 entries. when it belonged at the head of the standard series of multiplications taught in scribal schools.24 squares and inverse squares of integers and half-integers to 60. weight. and length systems respectively. consisting of the numbers only. students copied out a similar sequence of metrological units. We have just one or two administrative tablets with the sexagesimal workings still showing. first as indi- vidual sections and then as long extracts from the whole list. and practical calculation constants.21 and a handful of reciprocal tables.

Department of Archaeology. H. Philadelphia.} informal. Scribal students were expected not only to memorize arithmetical facts but also to put them into practice. 30. metrological and chronological tablets from the Temple library of Nippur. 1906. no. now in the Collection of the Babylonian Section. and only one other is headed. The student's copy is unfinished. from Nippur. but breaking columnar boundaries whenever necessitated by long names or explanations.4 A fifteen times multiplication 'table' for multiplicands 1-20. The tablet broke in antiquity. A S S Y R I A fig. (CBS 2142.28 As one might expect. 1. 40.32 T A B L E S I N SUMER. certain standard layouts . the more poorly executed student's copy in the right. Around a hundred published or partially published rough calculations and diagrams. show the students' numerical solutions to mathematical problems. Philadelphia. University Museum.27 Only one is ruled into more than two columns. from early second-millennium Nippur. Hilprecht. Mathematical. making a visual separation between the values and the names of the constants. The teacher's copy is in the left hand column. mostly on small round or square tablets from Ur and Nippur. and 50. University of Pennsylvania. B A B Y L O N I A . and parts of it are now missing. 94.

50 metrological lists. the sophisticated bureaucratic ability to display and manipulate numerical data was almost never exploited in the elementary training of future scribes. Larsa. Ur) and the 1640s (Sippar. c.The fact that this data is kid out in a landscape-oriented headed table. We can estimate the date of several groups of these tablets. TABLES IN SUMER. there was potential for innovations in data-management to cross over from administration into mathematics. 20 tabular calculations. so that it is very difficult to quantify the corpus. has gone completely unre- marked.29 Its fame derives from its mathematical content: fifteen rows of four extant columns containing sophisticated data relating to Pythagoras* theorem. At the moment we cannot even begin to date Babylonian tablets written in the later second millennium BCE. 300 standard arithmetical lists. Susa). c. or from clues gleaned from other tablets in the same museum accession lots. representing perhaps 5 or 10 per cent of such material in accessible museums. or give a good account of its chronological and geographical range. given the current clerical facility with tabular layouts. Ur. 50 non-standard ones. The earliest datable tablets are probably from the 1790s—80s (Larsa. Plimpton 322 has achieved such iconic status as the Mesopotamian mathematical tablet par excellence that it comes as quite a shock to realise how odd it is. Extraordinarily. then. around 430 exemplars have been published (c.BI. Shaduppum. despite their superabundance in archaeological museums. unless the scribes dated them for us.Yet there is only one known mathematical cuneiform tablet which is conspicuously indebted to administrative tabular practice. Clearly. of course. with two axes of organization and calculation. with a final heading MU. . A S S Y R I A 33 were used for particular problem types and students were encouraged to use columnar rulings to group data types. B A B Y L O N I A . 10 coefficient lists). The arithmetical and metrological lists could easily have been organized into multi-columnar headed tables. Nevertheless a few general statements can be made. are formal features of administrative tables from. c. They are known from almost all early second mil- lennium urban centres in and around southern Iraq. These. Larsa during the period of rigorous standardization in the 1790s—80s BCE. Nippur. it is puzzling that it happened so rarely. Relatively little contextual detail is known about the metrological and arithmetical lists and tables. c. and Uruk) but there are also groups from the 1750s-40s (Isin. either based on direct archaeological information. so badly do we understand the changes in writing conventions at that time. Because there is a general (mis-)perception that they are all identical they are heavily under-represented in publications. Mari.IM ('its name') for the non-numerical data.

Early publications often ascribe a Kassite date to multi- plication lists that we would now confidently date to the eighteenth cen- tury.15. 8g) to 100 talents (c. 1945. from Larsa. A large metrological Est has been discovered at Kanesh.30 while one of the coefficients lists may be from any time between fourteenth and the sixth century BCE. dating to c. 1. 10 litres) to 70 000 'ass-loads' (c. we cannot say with any confidence at all.33 There is only one list of Assyrian capacity and weight measures.34 TABLES I N SUMER.32 The earliest tablet from Ashur itself. Mathematical cuneiform texts.5 Plimpton 322. now in the University of Columbia Rare Book and Manuscript Library. 1850 BCE. It is now rather damaged but orig- inally ran from 1 shekel (c. an Assyrian trading post in eastern Turkey.31 Assyrian metrological and arithmetical lists and tables are equally thin on the ground. The heading on the first surviving column. Robson. Drawing by E. New Haven. There is still much speculation about the what the missing cotumn(s) on the left of the tablet might have contained. 3 metric tonnes). New York. gives important information about how the table was computed. Neugebauer and A. the only true table in Old Babylonian mathematics. 'The holding-square on the diagonal from which 1 is torn out. A S S Y R I A Fig. the extant part of which runs from 1 ban (c. in the absence of secure archaeological evidence. the complete set of multiplications from 50 to 1. 1200 BCE. Sachs. and squares of integers to 30. from Larsa c. 38-41. so that the short side comes up'.) Thus. which reads. whether any mathematical tables have survived from Kassite Babylonia. 1800 BCE. American Oriental Society. 0. c. pp. gives the standard reciprocals. (Plimpton 322. 7 million litres) and 2 shekels to just over . while the final column is simply a tine count. B A B Y L O N I A . in die classic early second-millennium format. The second and third columns of the table list the diagonals and short sides of 15 right triangles.

The tablet is signed at the bottom by one Bel-bani-apli ('the lord [god Marduk] is creator of an heir*). Thirteen of those tablets contain extracts from the standard list of weights and measures. where 'almost every room contained tablets which had been stored in large jars.38 The earliest are from a grovip of fifteen school tablets found in the archive of the city governor of Nippur. 755—730 BCE. of which a tiny number are mathematical or metrological. Nippur. A S S Y R I A 35 1 mina (c 16—500g). the later Assyrian capital: one enigmatic table of 3-place reciprocals in base 70.42 Tables for scholarship and astronomy While little use was made of tabular formatting for mathematical training. Whereas most mathematical lists and tables are known to us from several dozens.7cm) to 30 USH = 1 double-mile (c. B A B Y L O N I A . Uruk. 2.8 km). all from Ashurbanipal's palace library. or even thousands of exemplars. while the second has been disguised as a ration list by attaching a fictitious name to each entry and giving an (erroneous) total. the first is on the back of a letter. c. primarily syllabaries and religious texts.This suggests it was written by a professional scribe and designed as a reference document for a library. 59|.41 At the other end of the scholastic spectrum is a beautifully inscribed mathematical table. Ur. Ij. 10.36 and a few lexical tables giving the names of unit fractions in Sumerian and Akkadian. 1. from the early second millennium BCE onwards it became ubiquitous in other aspects of schooling and scholarship....'40 Sadly the exact disposition of the tablets within the jars was never recorded. inscribed on the same tablets as snippets from elementary syllabaries and word lists. none of which is truly tabular. It is the only identified table that combines both squares and square roots in a single entry. this document is unique. and Babylon itself. The table's horizontal axis of organization . 10-300 litres). 2% 3. whose square root is n\ where » = 1.. TABLES IN SUMER.34 A small table gives the ratios between length units from 6 grains = 1 finger (c. 59. The vertical lines of the tabular formatting appear to have served more as aids to produc- ing a beautifully written document than as conceptual separators of different classes of data. arranged round the room according to contents.35 There is even less from Nineveh. Sippar. Each line reads '« times n is tf.37 Several thousand school tablets of the first millennium BCE are also known from the Babylonian cities of Kish. hundreds. Both list capacity measures from 1 ban to \ gur (c.39 About a century later are the two hundred or so tablets from a badly pre- served building in Kish.

The first five lines of Proto-Ea go as follows: a2-a A 'The sign A can. U%. A S S Y R I A enabled inter-columnar relationships to be expressed between words. Outside Mesopotamia proper. Bant 'haruspex/extispicer/diviner'. be read as "aya" ia A The sign A can be read as "iya" du-ru A The sign A can be read "duru" e A The sign A can be read "e" a A The sign A can be read "a" '.36 TABLES I N SUMER. however. such lists also accrued Egyptian. phrases. or Ugaritic translations and could comprise up to five columns of text. A tiny percentage of urban Mesopotamians trained as scribes. Mesopotamian scholars carried one or more of the following titles: Tupsharru 'scribe/celestial diviner'. then and now they are known by their first line. or laments. with no headings.RA = hubullu (a type of loan). the scholarly corpus is so vast. B A B Y L O N I A . It is a real irony—and testament to our collective blindness to document formatting—that these intrinsically tab- ular documents are nowadays called 'lexical lists' while their list-like arith- metical counterparts are always known as 'multiplication tables'. Ashipu 'exorcist/healer-seer'.45 Each scholarly discipline had its own corpus of specialist literature. which enumerate the different Sumerian readings each cuneiform sign or sign-combination can take. From the early second millennium BCE onwards. Indeed. or longer texts. One of its first uses was in the school sign-lists now called Proto-Ea and Proto-Diri. Akkadian ummanu. attested from at least the eighteenth century BCE.43 An optional further column to the right gave one or more Akkadian transla- tions for each Sumeriati sign combination. prescriptions. complex. Human.44 Headings were never used. In the course of time the standard lists of Sumerian nouns also acquired columns of Akkadian translations. Experts in soothing angered gods. Each has a simple bi-columnar structure. and a tiny percentage of scribes became neither bureaucrats nor amanuenses but schol- arly experts. Hittite. Experts in extispicy [divination by the entrails of sacrificed goats] and lecanomancy [divination by the pat- terns of oil on water]. running to a hundred or more standard tablets of omens. Experts in interpreting celestial (and other) portents. Kalfi lamentation chanter'. Experts in curing diseases by drugs and physical remedies. Experts in magical manipulation of the supernatural. Am 'physician*. Tabular . and esoteric that only a tiny proportion of it has been published and studied thoroughly.

had thus become the temples.48 The series runs to 68 or 70 tablets (it exists in different versions) which were divided into four sections: lunar omens (tablets 1—22). solar omens (23—29). We. Most of our contextualized knowledge of the men themselves comes from three separate locales. which was found down a palace well with only a few fragments of its wax writing surviving. Parthians) with their own belief systems. however. for instance. will focus on the works of the celestial diviners—more commonly if anachronistically referred to in the modern literature as astronomers or astrologers. Greeks. ao. There is strong evidence to suggest that parts of it at least go back to the early second millennium. 'scribe of "When [the gods) Ann [and] Ellil" *. Until at least the second century BCE the full Akkadian title of the celes- tial diviner was tupshar EnumaAnu Ellil. B A B Y L O N I A .49 Fortunately the title and destination of the work had been inscribed in cuneiform on the front cover! Other copies found in Assyria had been looted from Babylonian libraries in 647 BCE. 200 BCE-50 CE and probably earlier). weather omens (30—49). Babylon was ruled by a succession of outsiders (Persians.d Ea. in their sure counsel had fixed the designs of heaven and earth. TABLES IN S U M E R .46 We also have tablets by and about the celestial diviners attached to two later Babylonian temples: the Resh of Ami in Uruk (c. A great many of their letters and reports survive. the great gods. who could no longer rely on state patronage. Ellil. whether. but the earliest surviving sources come from the Assyrian city of Kalhu. 750—612 BCE. At the Assyrian court at Nineveh. king of Assyria. which directly cite the scholarly reference works. for adding commentary to core works or for drawing up tables of propitious and unpropitious days of the month. c. and omens from stars and planets (50—70). they assigned to the hands of the great gods the duty to form the day well and to renew the month of mankind to behold. 400-100 BCE and perhaps later) and the Esangila of Marduk in Babylon (c. . a select inner circle of scholars wrote regularly to the king about what the future portended for matters of state and the well-being of the royal family. after the first line of the huge compendium of celestial omens that constituted their core reference work: When Anu. 720 BCE-—including a 'deluxe' edition written on 16 ivory writing boards for Sargon.47 By this late stage in Mesopotamian history. the professional locus of the scholars. c. thereby enabling us to see how they worked in practise. A S S Y R I A 37 documents were used sporadically throughout the corpus.

and annual Diaries.50 The first two tables have 30 entries each. They each end with one-line summary 'headings' but are otherwise very informal. the day watch is 2f mina. Duration of lunar visibility in the equinoctial month. that would last for around seven hundred years and even- tually (indirectly.51 All four model the moon's visibility using what are now called 'linear zigzag functions' whereby the variable increases or decreases by a constant amount over fixed intervals. list-like entries that are broken into sense units on the surface of the tablet. (D) In month Ayyar. [according to] the tradition of Babylon C. for instance. with verbose. is the third entry from each of the four tables: (A) On day 3 the moon is present for 15 (USH). Although these are the earliest known tabular representations of linear zigzag functions. monthly. hence their modern name. (B) On day 3: 1/2 niina 6 shekels. and in lunar visibility at full moon. as they are now called. Plotted on a graph over time such functions trace regular zigzags. the day 15: night watch is 2f mina. the schemes on which they are based can be traced right back to much shorter descriptions in mathematical and astronomical texts of the early second millennium BCE.52 At around the same time that Enuma Anu Eltil was being copied for the king of Assyria. and attains fixed maxima and minima. Seasonal variation in length of day and night D.38 T A B L E S I N SUMER. from at least 652 BCE. They formed a continuous tradition. the last two 24. Here. much of which is now lost. A S S Y R I A Only one ofEtiumaAnu Ellil's 70-odd tablets is tabular. These observations were collected into daily. Duration of lunar visibility in the equinoctial month. [according to] the tradition of Nippur B. B A B Y L O N I A . about a thousand years earlier. and in translation) comprised much of the observational . Tablet 14 in fact comprises four tables. scholars began to make systematic records of celestial phenomena without always taking omens from them. Monthly variation in lunar visibility and new moon. the moon's period of visibility day 1: is 10 USH. (C) In month Ayyar. all related to lunar motion: A.

direct and retrograde motion and stationary points. some time around 500 BCE. and occasional astronomical phenomena such as meteors and comets (including Haley's Comet in 164 and 87 BCE). [. For the planets they record dates and posi- tion among the stars of first and last visibility. and the date and description of lunar and solar eclipses. and length but more often reversed that order.The monthly Diaries typically contain the following data: A statement of the length of the preceding month. many of them come from Uruk and Babylon. They enabled scholars 'to devise consistent and powerful sets of rules (embodied in "Procedure Texts") that allowed them to generate tables (called by modern scholars "Ephemerides") permitting one to foretell the longitudes and times of the occurrences of lunar and planetary phenomena and the longitudes of these bodies in the intervals between these occurrences. the time interval between moonrise and sunrise on the morning of die moon's last visibility. columns continue from one side of the tablet to the other. however.53 These diaries were key to the development of mathematical astronomy. 18 years for the Moon). ] Weather condi- tions are regularly reported. TABLES IN SUMER. time intervals between sun/moonrise/set in the middle of the month. Frequently. where the celestial scholars were most active. the dates on which die moon approached the various Normal Stars and die watch of the night in which this occurred.. presumably to help scholars with their calculations.. 8 years for Venus.) The dates of die solstices and equinoxes and of the heliacal rising of Sirius are recorded.55 In many ways the dozen or so extant metrological lists and tables of the first millennium BCE are the inverse of their second-millennium counter- parts: they now followed the order of capacity. There are never any headings to these tables. Successive events are given in the cells going down a col- umn. Sometimes. the tablets turn sideways so that the rows continue from one side to the other.The similarities between events separated by these periods can then be easily viewed. more than one line) of a particular observed (or sometimes predicted) planetary or lunar event. B A B Y L O N I A . weight.g. A S S Y R I A 39 basis for Ptolemy's Almagest of the second century CE. [.Tables tended to put the sexagesimal (or more . and conjunctions with Normal Stars.. Not surprisingly. Some collections of lunar and planetary data are also set out in a ruled tabular format. area. the time interval between sunset and moonset on the first day of the month. Each cell of the table contains a description (often on.. and going along each row we get events separated by the character- istic planetary or lunar periods (e.*54 Alongside those Procedure Texts and Ephemerides increasingly accurate and sophisticated mathematical tables were developed.

II. They fall into two distinct groups: those for the Moon.40 T A B L E S I N SUMER. studied by 0. Providence. l.J. instead the small but growing corpus consists primarily of tables of many-place regular reciprocals between I and 3. S3. 1955. (ed. B A B Y L O N I A . I. III).57 It has yet to be shown whether these tables would be any use in calculating Ephemerides using the methods of the Procedure Texts. T. Late Babylonian astronomical and related texts. now housed in the British Museum. N. Berlin. recording lunar and solar eclipse possibilities for the years 12 ece-43 CE. V). about a third of known examples record the data for the new moon on the front of the tablet and the full moon on the reverse. ASSYRIA Rg. reprint edition. A. from Babylon. 115-16. Sachs). Neugeubauer. pp. no. Strassmaier. and those for the planets. and magnitude of eclipse (cols. (BM 34083 = Sp. date (here cols. at new and full moon. Ephemerides.) often decimal) equivalent to the left of the metrological units instead of the right. G.. 1955. no. were drawn up from at least 300 BCE until about 50 CE. Pinches and j. and many-place squares and square roots of integers and half- integers.S6Very few multiplication tables survive. from Babylon. always tabular. 49. IV). Two different calculational . 181.(> The latest datable Ephemeris. Rhode Island. longitude of Moon (col. It is a fragment of a larger tablet in which columns are grouped into threes. A typical lunar Ephemeris tabulates twelve to eighteen functions of the Moon over a whole year. Brown University Press. Astronomical cuneiform texts. Springer. 3 vote.

and System B. by drawing physical dividing lines between them. Perhaps it is no coincidence that tables took off only after the invention of the sexagesimal place value system and the concomitant conceptual separation of quantifier and quantified. The latest datable Ephemeris is from Babylon. Planetary Ephenierides were much less sophisticated. which is based on step functions for calculat- ing longitudes (in which a variable jumps discontinuously between constant values at fixed intervals). Conclusions Only rarely in Mesopotamia!! history were tables a mainstream document format. It was only with the development of mathematical . At that point their potential as powerful tools for the management of quantitative data was fulfilled remarkably rapidly. at least in and around the city of Nippur. TABLES IN SUMER. which uses linear zigzag functions of the sort used in EnumaAnu Ellil tablet 14/* Despite the complexity of the data they record. The material objects themselves facilitated conceptual advances in quantitative thinking. over a period of a few decades in the early eighteenth century. they account for only 1 or 2 per cent of all admin- istrative documents and scribes continued to prefer simpler linear or prosaic methods of managing information. an astronomical almanac predicting the main celestial events for the year 75 CE. the new format enabled numerical data and relationships to be seen and explored in ways hitherto unimaginable. B A B Y L O N I A . predicting lunar and solar eclipse possibilities for the period 12 BCE—43 CE. and this potential continued to be exploited (allowing for huge gaps in the evidential record) for fuEy five hundred years. In its turn. But even in the heyday of Mesopotamian tables in the first half of the eighteenth century BCB. A S S Y R I A 41 methods are used: System A. first appearing over half a mil- lennium after the invention of writing and only establishing themselves in the nineteenth century BCE. for Mesopotamian tables made manifest this distinction between quantitative and qualitative.The latest datable cuneiform record of them all is ako from Babylon. They took a long time to catch hold. cuneiform scribes became less ambitious in using tables to manipu- late and check numerical data as the writing board took over many of the functions of the tablet. Paradoxically. predicting just five or six key moments in the planets* journeys across the night skies. the columns of Ephemerides are never headed. even as tabular formatting began to infiltrate other textual types.

'The uses of mathematics . 241—60. pp. 1969. Archaic bookkeeping. and E. in Die Rolls der Astronomic in den Kulturen Mesopotamiens (ed. and lengthy variables of lunar and planetary theory. 21—9 and E. Graz. preferring to express arithmetical and metrological equivalences as lists.SfVeriagsgeselkchaft. Nemet-Nejat. Friberg. D. to appear. The exact sciences in antiquity.Journal of Near Eastern Studies 54 (1995). Further reading H. While the inner workings of Mesopotamian mathematical lists are generally well understood there is still much we do not grasp about their functioning within the larger educational system. and R. London and Chicago.Van De Mieroop. 1993. 1994. It also has lots of good photos. including the table of squares from Shuruppag. B A B Y L O N I A . H0yrup. H. in Tlie social and economic implications of accounting in the ancient world: a colloquium held at the British Museum. American Mathematical Monthly 109 (2002). Almost nothing has been written on administrative tables. A more recent summary is K. We understand even less of the history of truly tabular tables in Mesopotamia. K. Even more surprising is that Mesopotamian scholastic mathematics employed tables very rarely. Dover. Routledge. Damerow. in Companion encyclopedia oj the history and philosophy of the mathematical sciences (ed. New York. Neugebauer. 1993. Robson. Hudson and M. Robson. pp. A S S Y R I A astronomy in the latter half of the first millennium BCE that tables came into their own again to record and calculate the multiple. 383-405. Grattan-Guiness). I. November 2000. is an excellent overview of the development of writing and accounting in third-millennium Mesopotamia. Englund. 'On the structure of cuneiform mathematical table texts from the —1st millennium'. 3—52. The classic account of early second millennium mathematical tables is O. A survey of mathematical tables from later Mesopotamia is given in J. pp. complex.42 TABLES I N SUMER.'Words and pictures: new Eght on Plimpton 322'. their debt to administrative or astronomical practice is transparent. M. Where we do see truly tabular mathe- matical documents. Gaiter). For more general introductions to Mesopotamian mathematics. 2nd edn. P. see for the moment E. Robson. 'Systems for learning mathematics in Mesopotamian scribal schools'. ed. and their place in the history of ideas. 105-20. RM. London. University of Chicago Press. 'Mesopotamia!! mathematics*.Druck. 'Accounting for change: problematising the development of tabular accounts in early Mesopotamia*. see J. Nissen.

48. 1996. B. London. Freiburg and Gottingen. 44. Andrews University Press. no. M. A S S Y R I A 43 in ancient Iraq. H. 1984. while C. K. Neo-Sumerian account texts in the Horn Archaeological Museum. 1993. 2000. 215.998. E Walker. now in the Horn Archaeological Museum. Astronomy is discussed by J. R. 1993. Roafs Cultural atlas of Mesopotamia and the ancient Near East. P. 93—113. London and Chigago. 56. in Episodes in the early history oj astronomy. is a clear explanation of the writing and numeration systems of ancient Mesopotamia. UCLA. 58-9. London. The cuneiform digital library initiative. Archaic bookkeeping. 'Texts from the Uruk period". 2001. Notes 1. Michigan. See. in The organization of power: aspects of bureaucracy in the ancient Near East. M. Civil.edu/progress. Kluwer. Attinger and C. 36—46. M. University of Chicago Press.edu/pioposal. H. . no. Heidelberg. CDL Press. 1990. Andrews University Press. 9. Englund. 6. Fitzgerald. 1—23) sets out the observational basics. Archaic book- keeping. pp. Drehem. 6000—600 BCE*.'Texts from the Umk period*. Discussed in detail by R. Sigrist. K. M. Dordrecht. 2. Nissen.The first chapter of the same book. especially p. Freiburg and Gottingen. R. in Spatumk-Zeit und Frithdynastische Zeit (ed. <http://www. Sigrist. 'CDLI proposal'. 1998. probably from Puzrish-Dagan. Biggs). McG. Sigrist. in Astronomy before the telescope (ed.cdli. 7. P. E Walker). 2001. Englund. 2001. 17 on pp. M. Britton and C. Berrien Springs. Springer Verlag. 56—64. K. Selin). 2nd edn (ed. pp. British Museum Press. Englund. Berrien Springs. Michigan. probably from Puzrish-Dagan. pp. AUAM 73. K. pp. D. in Mathematics across cultures: the history of non- Western mathetnatics (ed. Englund. 35—44. B A B Y L O N I A . 1. Chicago. Aaboe. Englund. Damerow. '["he cuneiform digital library initiative. P.'What every young person ought to know about naked-eye astronomy' (pp. in Spatumk-Zeit und Fruhdynastische Zeit (ed. R.0693. 'Ur III catalogue'. 5. M. esp. University of Chicago Press.992. 1992.>.'Astronomy and astrol- ogy in Mesopotamia'. Attinger and C. p. C. E Walker's Cuneiform. University of Chicago Press. et aL. 4. Nissen. quote from p. AUAM 73. 'Babylonian arithmetical astronomy'. Bethesda. pp. Facts on File. Neo-Sumerian account texts in the Horn Archaeological Museum. 8. Gibson and R. Uelinger). and R. pp.pdf>. British Museum Press. TABLES IN SUMER. 15—233. is an excellent overview of Mesopotamia!! history. 1. fig.0400. pp. 42—67 and by A. K. K.uck. B. Uelinger). 3. London and Chigago. <http://www. for instance the accounts of the 'Kushim' brewing archive described in H. 30. 1987. 1984. and R. 'Ur III bureaucracy: quantitative aspects'.ucla.cdli. UCLA. now in the Horn Archaeological Museum. Englund. B. P. Damerow. New York. 24-65.htnil.

1988. vol. 53-60. Sigrist. H. O. 22. 'HS 201—Eine Reziproken-tabelle der Ur HI-Zeit'. Malibu. Satgonk inscriptions from Adab. 68—72 (metrology). now in the Hilprecht Sammlung.J.Van De Mieroop). University of Chicago Press.44 TABLES I N SUMER. Historia Mathematica 3 (1976). p. D. vol. Berlin. 1906—1912. 18. Oelsner. in The social and economic implications of accounting in the ancient world: a colloquium held at the British Museum. 1935. Postgate. from Adab. O. A. Helsinki. 70. 12.g. Changchun.. from Nippur. no. M. Documents from the temple archives of Nippur dated in the reigns of the. Neukirchen- Vluyn. in Ushan mithurti. A S S Y R I A 10. HI Fiche 2.. 70)'. vol. Damerow). E. p. E. American Oriental Society. Powell. 10. IHAC. A 681. Clay. A more detailed presentation of this discussion is in E. Inscriptions from Adab. 14. BM 106425 and BM 106444 from Umma. 13. 2. Hudson and M. D. 3 vols. 11. Chicago. Verlag Butzon. to appear. Akademie Verlag. The University Museum. 101-4. Springer.Wiseman. University of Jena: J. 1969. N.'Assyrian writing boards'. Clay. . O. Robson. pp. M. 1. Berlin. 3.'Eine altsumerische Rechentafel (OIP 14. 417—39. B.T. 8—14 (reciprocals). now in the Oriental Institute. Mathematische Keihchrift-texte. H. A. vol. Sozial-okonomische Jexle und Rechtsurkunden aus Nippur zur Kassitenzeit. November 2000 (ed. Chris Martin. Martin. a colloquium held at the British Museum. Les sattukku dans I'Eshumesha durant la periods d'Isin et Larsa. 36. Bemhardt. pers. and C. 'Accounting for change: prohlematising the development of tabular accounts in early Mesopotamia'. from Girsu. cotnm. 1988. place notation and the early history of Babylonian mathematics'. Hudson and M. in The social and economic implications of accounting in the ancient world'. 1989. 202. J. p. 1. p. Helsinki University Press. 1. 1976. now in the British Museum. Lafont. 36—42 (multiplications). Mathematische Keilschrift-texte. Mathematical cuneiform texts. Berlin. Neugebauer. Iraq 17 (1955). Yang Zhi. N. pp. Heyrup and P. Birmingham. Undena. Robson. Dietrich Reimer Verlag. 2001. Philadelphia. I. p. F. Neugebauer and A. Sachs. Springer. M. 15. 1984. Philadelphia. D. to appear. The University Museum. Chicago. 1st T 7375. M. pp. 1906. pp. Fara. A. pp. 5. pp. Fara.T. Luckenbill. Fales and J. 23. 16. pp.'Accounting for change: problematising the development of tabu- lar accounts in early Mesopotamia*. 21. 1930. Edzard . B A B Y L O N I A . Cassite rulers. vol. Appx. now in the Istanbul Archaeological Museum: O. Imperial administrative records. Proust). Birmingham. 82-103. 49-50 (reciprocals and multiplications). New Haven. 14. Neugebauer. and other tablets from Umma Nippur.'The antecedents of Old Babylonian. 2 vols. in Changing views on ancient Near Eastern mathematics (ed. 17. 1935. Documents from the temple archives of Nippur dated in the reigns of the Cassite rulers. 3—13. 19. and Girsu (unpublished. Chris Martin. Koslova. 1992-95. Martin. Festschrift Wolfram Freiherr von Sodm (ed.Van De Mieroop). November 2000 (ed. M. D. 20. W Rollig). R. Berlin. HS 201.

11. Berlin. Neugebauer. vol. no. 1945. American Oriental Society. Friberg. pp. 1. pp. 10—13 and passim. 24. 1-40. Mesopotamian mathematics. pp. 2100-1600 BC. Berlin. 29. Clarendon Press. 2100—1600 BC. Oriental Institute. p. pp. Edzard).j. in Die Rolle derAstronomic in den Kulturen Mesopotamiens (ed H. O. Graz. D.V. 1935. in Reallexikon der Assyriologie. TABLES IN SUMER. pp. Mesopotamian mathematics. vol. 1990. 34. VAT 9840. Friberg. Oxford 1999. 7 (ed. Mathematische Keilschrift-texte. in Reallexikon derAssyriologie. Michel. Springer. 549-50 [in English]. 7 (ed. 70. Current whereabouts unknown. Berlin. Robson. Hinrichs. 33—4. Keilschrift-texte ausAssur: verschiedenen Inhalts. 1945. 26. 2100—1600 BC. Philadelphia 1906. 25. Gaiter). D. Berlin. in Reallexikon der Assyriologie. Oxford. in Die Rolle der Astronomic in . Mathematical cuneiform texts. Berlin. Sachs. 1. E. from Ashur.'Neither Sherlock Holmes nor Babylon: a reassessment of Plimpton 322*. Schroeder. Springer. E. 30. Leipzig. Edzard). 'Un petit texte d'Assour*.es Museum. now in the Vorderasiatiscb. 1935. 68—72. pp. 73—5. 27. J. D. O. Mesopotamian mathematics. 1. Neugebauer. American Mathematical Monthly 109 (2002). J. 1990. 1993. 546-7 [in English]. O. Prosecky). 31. now in the Istanbul Archaeological Museum and the Vorderasiatisches Museum.Brack. Mathematische Keilschrift-texte.'On the structure of cuneiform mathematical table texts from the —1st millennium'. pp. pp. O. J. 1990. 28 (2001). WaJther de Gruyter.Walther de Gruyter. New Haven. American Oriental Society. Springer. D. 7 (ed. Springer.vol. O. 11-13 and 245-77. B A B Y L O N I A .'Words and pictures: new light on Plimpton 322'. List B. Walther de Gruyter. 1990. Oxford. Clarendon Press. O. metrological and chronological tablets from the temple library of Nippur. followed by O. 1935. 105-20. 542-6 [in English]. vol. Neugebauer. C. vol.Berlin. 546-7 [in English]. pp. Historic Mathematica. Edzard). D. O. 19—33 (multiplications). 46—7.& Verlagsgeselkchaft. 11—12 (reciprocals). 1920. 1st A 20 + VAT 9734. Edzard). vol. Mathematical cuneiform texts. J. p.J. Clarendon Press. Robson. vol. Sachs. 34. pp. pp. pp. 33—5. Hilprechf. p. E Thureau-Dangin. Mathematische Keilschrift-texte. E. 35. Berlin. Mathematical. 26. 'Mathematik'. Friberg. Prague.'Mathematik'. 32. 4—6 (metrology). For instance H. O. 1999. Friberg. J. 1 (ed. 1998. Berlin. Robson. Mathematische Keilschrift-texte. New Haven. 249-67.'Mathematik'. Neugebauer. 'Mathematik'. O. 1999. Walther de Gruyter. J. RM. University of Pennsylvania. Berlin. 184. 388. Friberg. Neugebauer and A. Friberg.'On the structure of cuneiform mathematical table texts from the —1st millennium'. 'Les tnarchands et les nombres: Fexemple des Assyriens a Kanish'. pp. Berlin. C. Neugebauer and A. 1935. 33. 1. pp. in Intellectual life of the ancient Near East (ed. in Reallexikon der Assyriologie. Robson. 28. Robson. A S S Y R I A 45 1945. vol. O. from Ashur. Revue d'Assyriologie 23 (1926). E. E.

vol. K. 2000. Beaulieu. Hunger. Neumann). Cole. Miinster. Oxford. 48. Oxford. vol. P. 'Scribes and scholars.'Ancient Mesopotamia!] lexicography'. Ash 1924. pp. Marzahn and H. 1. Springer. S. 1923-1933. 45. 30. Scribner's. Berlin. and D. Civil. K. Gaiter). 41.Druck. nos. 33. S. 255. 2000. RM. M. pp.796. R. K. 2000. Groningen. RM. Nimrud: an Assyrian imperial city revealed. Gaiter). Ugarit-Veriag. with their forerunners and related texts.-A. 1935. See E Rochberg. Berlin. Helsinki.E. 1993. Ugarit-Veriag. See in general Chapter 1. Civil. See M. Brown. now in the British Museum.123. pp. Springer. in Assyrioiogica et Semitica: Festschrift jur Joachim Oelsner (ed. O. Biblical Institute Press. Rochberg. Schuluntenicht in Bahylonien int ersten Jahrtausend v. AaA — naqu.Druck. pp.979. J. W. 1992 and S. D. Munster. P. 34. D. in Assyrioiogica et Semitica: Festschrift fitr Joachim Oehner (ed. Helsinki. in Die Rolle der Astronomic in den Kulturen Mesopotamiem (ed. Helsinki University Press. 4 vok (ed. Clarendon Press. P. 2069. 1978. . J. Munster. Parpola.& Verfagsgesellschaft. 'The astronomer-astrologers-—the scholars'. 1935. Ugarit- Verlag. 1. Astrological reports to Assyrian kings.W. vol. F. British School of Archaeology in Iraq. D. 1—16. 10. pp. 30-52. 2001. The letters and reports themselves are published in English translation by H. Letters from Assyrian and Babylonian scholars. from Nineveh. 40. 97—9. Groningen. A S S Y R I A den Kulturen Mesopntamiens (ed. pp. Oates. H. Graz. pp. Marzahn and H. 2000. Sasson). vol. P. Styx. R. p. 36. in the Ashmolean Museum. 124. M. 37-8.& Verlagsgesellschaft. now in the British Museum. B A B Y L O N I A . pp. the lupshar Enuma Anu EntiF. 39. I. B1 and B: with miscellaneous lexicographical texts jrom the Herbert Weld Collection. 60. text 4. 2001. Green. pp. Now in the Ashmolean Museum. Helsinki University Press. 10. 1935. Neugebauer. M. K. State Archives of Assyria. Gesche. nos. 1996. p. Brown. J.128.. The descendants of Sin- lecp-unninni'. Kish excavations. Chicago. Rome. 56. Styx. 38. and W. See most recently J. London. R. Lambert. Mathetrutische Keibclwift-texte. 42. G. 2. Nippur IV: The early Neo-Babylonian governor's archive from Nippur. Mesopotamian planetary astronomy-astrology.'The cultural locus of astronomy in Late Babylonia*. 8687. Graz. 31—46. pi. 49. Mathematiscke Keilschrift-texte. 46. Neumann). Mesopotamian planetary astronomy-astrology. EaA = naqu. Springer. 30-3 & pi. Neugebauer. Berlin. 359-75. H. 44. 47. See most recently D. 2000. Moorey. State Archives of Assyria. (earn Nineveh. vol. O. O. Syllabaries A. 43. 37. 34 [W 1931-38]. Gxoningen. 28-9. 89. 8. Clarendon Press. Chi. remainder unpublished. D. Styx. 1993. 1992. in Civilizations of the ancient Near East.75. Oxford.46 T A B L E S I N SUMER. R. p. p. van der Meer. Neugebauer. Brown. Mesopotamian planetary astronomy-astrology. 50. 2305—14. 1938. Mathematische Keikchrift-texte. New York. University of Chicago Press.

Lists of published mathematical tables from the later first millennium BOB are given by J. H. Britcon and C. 347-555. Britton and C. p. H..Walker). 53. British Museum Press. H. Walker). D. Mass. 52. B. Astronomical diaries and related texts from Babylonia. R. 1999. Osterreichischen Akademie der Wissenschaften. 55. as well as for detailed comments on the rest of this section. and A history of ancient mathematical astronomy. 51. 71-87 andj. Vienna. Osterreichischen Akademie der Wissenschaften. Springer Verlag. B. Swerdlow). Sachs and H. R. J. 62. 1955. TABLES IN SUMER. 'On the structure of cuneiform metrological texts from the —1st millennium'. Friberg. Springer Verlag. Heidelberg. 54. 130-48. Cambridge. in Die Rolle der Astronomic in den Kulturen Mesopotamiens (ed. Friberg.The Normal Stars were a group of thirty-one stars in the zodiac belt. J. Steele. MIT Press. F. The Diaries are pub- lished by A.. Sachs. 1996. and C. 383-405. SO. B. Fermor. Hunger and D. pp. 52-73. F. 212. C. 1996. Gaiter). Walker. Aaboe. Pingree. 1 tnina = 4 hours (weight running through a water-clock). 179-186. J. 'A new mathematical text from the astronomi- cal archive in Babylon: BM 36849'. M. Heidelberg.. pp. Astronomical cuneiform texts. in Astronomy before the telescope (ed. London. A. See D. 3 vols. Astral sciences in Mesopotamia.Druck. RM. 5. vol. pp. E Walker. Hunger. Hunger. Leiden. B A B Y L O N I A . p. B. Baghdader Mitteilungen 28 (1997). 'Astronomy and astrology in Mesopotamia*. George. E Walker. N. Astronomical diaries and related texts jmm Babylonia. Tablets of this type are published by H. 58. 'Enuma Aim Enlil XIV and other early astronomical tables'. . Archivfur Omntfomhung 38-39 (1991-92). Neugebauer. B. Jupiter. 1996. Mars. P. 251—365. 'Astronomy and astrology in Mesopotamia'. I thank John Steele for this information. p. 1 USH = 4 minutes (1 degree of arc).The planets observed were Venus. used as reference points for die movements of the moon and planets.&Verlagsgesellschaft. 1975. and J. J. Vienna. 'A table of 4th powers and related texts from Seleueid 'Babylon'Journal of Cuneiform Studies 43-45 (1991—93). E N. British Museum Press. 57. London.'Seed and reeds continued: another metro-mathematical topic text from Late Babylonian Uruk'. 'The water-clock in Mesopotamia1. Al-Rawi and A.The multiplication tables are published by A. Graz. 2001. A S S Y R I A 47 50. 3 vols. Arehivjur Orientfomhung 46-47 (1999-2000). Brown. Britton. vol. in Ancient astronomy and celestial divination (ed.. 1993. M. and Saturn. C. Brill. Procedure Texts and Ephemerides are edited and discussed by O. F. 1988-96. 56. in Astronomy before the telescope (ed. 1. Mercury.

.

Book i describes the ideas of arithmetical and geometrical progressions and the definition of the logarithm of a sine. The page shown opposite gives. the sines and logarithm of sines respectively of the angles 2° 0' to 2° 30' and. By the time that Briggs's Trigon&metria Britannica Fig. and the differential for the logarithm of the tangent. . printed by Andrew Hart in Edinburgh in 1614. in the fifth and sixth columns. The book continues with a demonstration of the properties of logarithms with regard to proportional numbers and the extraction of square and cube roots. in the second and third columns. contains a dedication to Charles. 2. The term ant/logarithm is used for what is now called the logarithm of the sine. together with some laudatory verses. 2 The making of logarithm tables GRAHAM JAGGER In 1614 the invention of logarithms burst upon the world with the publica- tion of Napier's Mirifici logarithmorum canonis descriptio. or A description of the admirable table of logarithms. in steps of one minute.1 Napier's Mirifici Logarithmorum canonis descriptio. There follow 90 pages of tables. Prince of Wales. the logarithms of sines and cosines respectively of the angles 87° 30' to 88°.

known as 'Napier's bones*. Edinburgh. Salvators College. which we examine here. Briggs's Logarithmorum chilias prima (1617) was the first published table of logarithms of numbers to base 10. This process of maturation. Napier continued his studies abroad although it is not known where or what he studied. the eighth Laird of Merchiston. Napier's rather clumsy ini- tial formulation had been transformed into one of supreme elegance. a work which described a number of calculating devices including the method of multiplication and division by numbering rods. was born in Edinburgh in 1550. James VI (in 1603 to become James I ofEngland) to see that'justice be done against enemies of Gods church* and to 'purge his house. Andrews. and. It seems that after an initial education at St. . Janet Both well. each face carrying one of the multiplication tables from 1 to 9. St. In 1617 he published his Rabdologiae. Certainly by 1571 he had returned to Scotland and was married there in 1572. Henry Briggs and Edmund Gunter. was marked in particular by the work of two other British mathematicians. Napier's interest in aids to calculation was not confined to logarithms. as well as English. John Napier and his logarithms John Napier. and German.2 These consisted of rectangu- lar rods of square cross-section. the son of Sir Archibald Napier and his wife. A plaine discovery went through several editions in Dutch.1 Napier first gained a reputation as a scholar and theologian with the pub- lication of his A plains discovery of the whole revelation of Saint John in 1593. Atheists and Newtrals'. followed within three years by Gunter's Canon triangulorum (1620). the first table of the logarithms to base 10 of trigonometrical functions. Two additional rods were devised to aid the calculation of square and cube roots. French. where he lived until his death on 4 April 1617. wood. even precious metals. On the death of his father in 1608 Napier moved to Merchiston Castle. family and court of all Papists. This work of Biblical exegesis arose out of the fears entertained in Scotland of an invasion by Philip II of Spain. daughter of an Edinburgh burgess. ivory.50 THE M A K I N 6 Of LOGARITHM TABLES appeared in 1633 logarithms had come of age. These rods were popular for many years after Napier's death and were sold in a variety of forms made of paper. the utility of which was to remain unsurpassed until the invention of the elec- tronic calculator. as objets de salon. and in it Napier urged the Scottish king.

about 105 cm in height. tangents and secants. . Edinburgh. of a Victorian representation found on the exterior of the Scottish National Portrait Gallery.. Napier describes a more complicated system of engraved rods or strips for mechanical calculation. Divisions can also be done by its use. it deserves a better place than to be put last in the book. THE M A K I N G OF L O G A R I T H M TABLES 51 Rg. In an appendix to the Rabdologiae. or the tables in. can be done readily and with maximum speed.. any multiplication. This is a copy. this work. By using it. The appendix begins:3 Although tills promptuary was discovered by me least of all. The four chapters which follow the introduction describe the construction and use of the promptuary. headed 'The Promptuary for Lightning Multiplication'.2 Statue of John Napier. but they first have to be converted into multiplications. no matter how difficult or involved. using either tables of sines. 2.

was published posthumously by his son Robert in 1619.5 From the very beginning of logarithms their utility to navigators has been of supreme importance in their development.4 Kepler had once worked inTycho Brahe's observatory and so was able to secure information from Brahe directly. was published. was to explain in detail the way in which the tables in the Descriptio had been calculated and the rationale behind them. the degree to which it relates being given at the top left hand corner of the page. Napier set out to construct a table by which multiplication of these sines could be replaced by addition: the tables in the Descriptio are of logarithms o( sines. especially sines. Napier's second work on logarithms. surpasses the method of Napier.52 THE MAKIN6 OF L O G A R I T H M TABLES But Napier's greatest contribution to mathematics was his logarithms and it is to a description of these that we now turn. Napier's Descriptio Realising that astronomical and navigational calculations involved primarily trigonometrical functions. to translate it into English for the benefit of the Company's seafarers. Kepler comments on the state of trigonometry and then goes on to say: 'But nothing. which comprises material written by Napier many years before. Each page covers thirty . even in the year 1594. Mirifia logarithmorum canonis descriptio (known as the Descriptio). The object of this latter work. Only the minute is shown. There is some evidence to suggest that Napier began his work on logarithms at least as early as 1594. Mirifici logarithmorum canonis constmctio (the Consfructio). The East India Company was so impressed by Napier's Descriptio that it asked Edward Wright. the first column contains the angle increasing from 0° to 45° (the semi-quadrant) in steps of 1 minute. That is. and are semi-quadrantally arranged. the Descriptio employs 'logarithms* even though this latter term was a later invention. They consist of seven columns. In a letter written in 1624 to his friend Cugerus. although a certain Scotchman. held out some promise of the wonderful Canon in a letter toTycho*.6 But whereas the Constmctio uses the term "artificial numbers'. in my mind. It was not until 1614 that Napier's first work on this subject. a Cambridge mathematician and expert in naviga- tion. In addition to tables of logarithms the Descriptio also contains an account of the nature of logarithms and a number of examples explaining their use.

Rather he would have taken it as half the length of the chord subtended by twice the given angle at the centre of a circle of arbitrary radius. also in steps of 1 minute. the trigonometric function that is equal in a right-angled triangle to the ratio of the side opposite the given angle to the hypotenuse. called the whole sine because it was the maximum value that the sine could reach. an arrangement following that introduced by Rheticus in his Canon doctrinae triangulomm of 1551. in columns 2 to 6. The seventh column contains the complementary angle decreasing from 90° to 45°. and the third and fifth columns. the logarithm of its tangent. the logarithm of . minutes. and avoid fractions. Thus for a given angle in the first column we have essentially. to whatever accuracy was desired. headed Sinus. headed Logarithmi. THE MAKING OF LOGARITHM TABLES 53 The definition of angular functions at the beginning of ttie seventeenth century The modern definition of the sine of an angle. the logarithms of their respective sines. its sine. was chosen to be sufficiently large to enable all the quantities to be considered as whole numbers. Napier chose a value of 10000 000. contains the differences between these two logarithms. headed Differentiae. the logarithm of its sine.7 The second and sixth columns. This radius. the table thus occupying 90 pages. would have been unknown to Napier. The fourth column. contain the sines of the angles in the first and seventh columns respectively.

In keeping with the definition of angular functions then current.Johns College. in their original form. but thereafter it was to become Gresham College. respectively. The second difficulty. in the parish of Halifax. but the complete adaptation to the decimal scale also required the choice of the log- arithm of unity as zero. went to St. though.This arises direcdy from Napier's definition of a logarithm. In July 1575 Sir Thomas Gresham. log 10. this time counting the columns from right to left. make them dif- ficult to use. Seven subjects. one on each day of the week. probably in early 1561. is that these logarithms are not well adapted to decimal arithmetic. save that column 4 now contains the logarithm of the cotangent. the values in the table are integers in the range 0 to 10000000.8 The substitu- tion of a power of ten for this number would resolve this difficulty. divinity. astronomy. were to be offered. Yorkshire. the value of log l. medicine.54 THE MAKIN6 OF L O G A R I T H M TABLES its cosine and its cosine. Cambridge. where he took his BA in 1581 and MA in 1585. He was made a Fellow in 1538. The same rule hold for angles in the right-most column. Two features of Napier's logarithms. In 1596 Briggs became the first professor of geometry at Gresham College in London.£50 and with an official residence in the College.. which follows directly from the first. music. and rhetoric. For each subject an unmarried professor was to be appointed at an annual salary of . that log be/a = log b + log c — log a. since the multiplication or division of a number by 10 corresponds to the addition or subtraction of another unwieldy number 23 025 842. law. which was . the value of the whole sine.9 He was educated at the local grammar school and. willed his house in Bishopsgate Street to his wife for her lifetime. an institution where free lectures were to be provided for all who cared to hear them. in 1577. There are no separators of the digits. geometry. Henry Briggs and the Logarithmorum chilias prima Napier's invention of logarithms was rapidly and enthusiastically taken up by Henry Briggs. Briggs was born at Warley Wood. but also the manipulation of the number 161 180 956. the founder of the Royal Exchange. Note. namely that log ab is log a + log b — log 1 and log a/b is log a ~~~ log b + log 1. Gresham College. It was these two apparent deficiencies which Napier and Briggs later sought to address. The first is that the calculation of the logarithm of a product or a quotient involves not only the addition or subtraction of two numbers.

2. who was a friend of Briggs. and Briggs secured the election of Henry Gellibrand to succeed Gunter in 1627 Fig. the Roya! Society's first home. . merchant and founder of the Royal Exchange. that Gresham. and he took up his duties at Merton College in January 1620. a post that he held until his death on 26 January 1630. was inaugurated as a seat of learning in 1597 as part of a bequest made by Sir Thomas Gresham (1519-1579). The College still exists as an educational institution but the original building. was demolished in the eighteenth century.3 Gresham College. THE MAKING OF LOGARITHM TABLES 55 to play a crucial role in the early development of logarithms. Edmund Gunter. on the site now occupied by Tower 42 (the former NatWest Tower). formerly Gresham's mansion in the City of London. It is mainly due to the work and influence of Henry Briggs. The early years of Gresham College were plagued by disputes between the professors and the trustees. came into existence after the death of Lady Gresham in 1586 and many legal squabbles. became professor of astronomy in 1620. to the south of the modern Liverpool Street Station. College has any claim to scientific respectability at all in these years. It stood in Bishopsgate.10 Briggs was later invited by Sir Henry Savile to become professor of geometry at Oxford. who was the professor of geometry at the College from its foundation until his move to Oxford in 1620.

for I never saw book. which pleased me better. After some discussion. and Gellibrand ensured that for its first 40 years Gresham College had a local and practical relevance. During Briggs's early years at Gresham College he worked on problems of navigation. was published by Blundeville in 1602. I hope to see him this summer. though. seconds.17 In the summer of 1615 Briggs visited Napier in Edinburgh. It was within this group that the developments in theory. 5 degrees 44 minutes and 21. hath set my head and hands a work with his new and admirable logarithms. lord of Markinston.56 THE M A K I N 6 OF L O G A R I T H M TABLES after the latter's death in 1626. and at once set about trying to rectify them. it emerged that Napier'. Between them Briggs. if it please God. He was. when expounding [the] doctrine [of Napier's logarithms] publicly in London to my auditors in Gresham College. from 1610. together with a description by Gilbert of two magnets which he had invented. Gunter. later to become archbishop of Armagh.*2 Among his contemporaries Briggs was not alone in his inter- est in navigation.16 should be 10000000000. that is to say. Briggs was immersed in logarithms. and instrumentation took place which were to transform the art of navigation into a science. however. taking with him some tables that he had prepared along these lines.. On 10 March 1615 he wrote to Ussher:'4 Naper. computational techniques.. In an age of expanding trade and colonization. And con- cerning that matter I wrote immediately to the author himself. This table.11 and between 1608 and 1612 he collaborated with Edward Wright in compiling observational tables for a new edition ofWright's Certaine errors in navigation. aware of the cumbersome feat- ures of Napier's canon. [I] remarked that it would be much more con- venient that 0 should be kept for the logarithm of the whole sine (as in the Canon mirifaus) but that the logarithm of the tenth part of the same whole sine. and made me more wonder. and in one of his letters. was of opinion that the change should be effected . the magnetical declination being given. Briggs later recalled that . Briggs became an ardent publicist for logarithms and spent the rest of his life in their development. the coun- try which could develop new navigational techniques had a competitive advantage over its rivals in the quest for new colonies and their attendant natural resources..15 which he regarded as defects.. writing A table to find the height oj the pole.13 By 1609 Briggs was in correspondence with James Ussher. Briggs tells Ussher that he has been studying eclipses. By 1615.

and the second. By the time of his second visit to Napier in the summer of 1616. Edinburgh. the first point moving uniformly and the second point moving so that its velocity is everywhere proportional to its distance from S. 'John Napier. T' Z.Whiteside ('Mathematical thought in the later 17th century*. Imagine two straight lines. GS is the sine and TS the whole sine. p. Briggs had already composed the 'principal part' of what was to become the Arithmetica logarithmica of 1624.Macdonald). 19. TS of fixed length. lltf mmtnutioti of rite tiviiiterfitl cation of iogtiritlttns (translated by WR. It follows from Napier's definition that his logarithms of sines decrease monotonically as the angle. . the first. 1889. Briggs would have visited him again in the summer of 1617 had Napier not died in April of that year."'' Napier's definition can be simply stated.Btackwood. Briggs's account that he and Napier rapidly arrived at a consensus and that the invention of logarithms to base 10 was a joint accomplishment. or radius. Archive for history of exact sciences 1 (I960). p. If the points reach L and G respectively at the same instant then the number that measures the line T'L is defined as the logarithm of GS. Napier's distance-speed model is medieval rather than modern and its now usual treatment by methods of the calculus is quite foreign to its kinematic nature. 220) has pointed out.'17 Briggs readily agreed that this was the best way of pro- ceeding and began at once to calculate some logarithms on this basis. in this manner.T. of infinite length. THE M A K I N G OF L O G A R I T H M T A B L E S 57 Napier's definition of logarithms Central to Napier's definition of logarithms was his use of the concept of motion. increases. at the same instant and with the same velocities. and therefore the sine. of points moving along lines with speeds defined in various ways. Two points start from T" and T respectively. As D. that 0 should be the logarithm of unity and 10000000000 that of the whole sine. It is clear from.

He [Briggs] not unwillingly gratified him [Napier]. together with the appended logarithms. These. which I presume he has sent to you. can be used.'23 Unlike Napier's logarithms. not with the intention of becoming public property. it is clear from its content that the author could only have been Briggs. and the logarithm of that number in another. as soon as conveniently may be. In this scheme. God willing.'for the calculation of all triangles'.21 it is to be hoped that his posthumous book will shortly give us abundant satisfaction. or after December 1617. so called from the first words of the preface to the Latin original. proportional to the actual sines instead of indi- vidual degrees and hundredths of degrees:20 this he hopes to publish.22 He [Napier] did not cease from urging the author [Briggs] (when he twice visited him at his home in Edinburgh and. since it is undoubtedly the work referred to in a letter from Sir Henry Bourchier to Ussher dated 6 December 1617 in which Bourchier says 'Our kind friend. accurately drawn up from first elements by himself ten years earlier. through algebraic equations and differences. hath lately published a supplement to the most excellent tables of logarithms. and had shown him the first part of these Logarithms which he had then solved) to take this work upon himself. In a slight volume neither the enjoyment nor the toil has been slight. together with his earlier table of sines. of ever-to-be-revered memory. most kindly welcomed. . published in his Wonderful Table. the Logarithmotum chilias prima. For he has the Table of Sines. those of Briggs were the logarithms of the first thousand natural numbers. The reference to Napier's death shows that it could not have been written before April 1617. but partly to satisfy on a private basis the wish of some of his own intimate friends: partly so that with its help he might more conveniently solve not only several following thousands but also the integral table of Logarithms used for the calculation of all triangles. had gladly stayed -with him for several weeks. Mr Briggs. The calculations which Briggs had in mind here are those involved in finding all the sides and angles of a triangle given that at least three of the six dimensions (three sides and three angles) of the triangle are known. But because diese logarithms are different from those which the most renowned inventor [Napier]. This is an extremely rare book of only 16 pages and is the earliest printed table to base 10.58 THE M A K I N 6 OF L O G A R I T H M TABLES Shortly after Napier's death Briggs privately printed the 'principal part* of his logarithms. says Briggs. the determination of the logarithm of a sine would involve first looking up the sine in one table.18 It has no title-page but begins with a Latin preface of which the following is a translation:19 Here is the first thousand logarithms which the author has had printed. Although the preface is not signed.

the line separating the characteristic from the mantissa has been ruled in by hand. the achievement of Edmund Gunter.25 The logarithms of Briggs2* were of integers: a base 10 equivalent of Napier's logarithms of sines was not to appear until 1620. as in the page shown. have been written in and the line separating the characteristic (the integral part of the logarithm) from the mantissa (the fractional part) has been ruled in by hand. The right hand side of each page was originally blank but in one of the British Library copies24 some of the differences.4 A page from Briggs's Logarithmorum chllias prima. This rare work is a smali pamphlet of only 16 pages and is the earliest printed table of logarithms to base 10. The digits are separated into groups of five by commas. 2. The right hand side of each page was originally blank but in the British Library copies some of the differences have been written in. 1617. . to four figures. Fig. and. THE M A K I N G OF L O G A R I T H M T A B L E S 59 The logarithms are given to 14 figures.

. and sent for Briggs from Cambridge. of his sector (which came to be called 'Gunter's Sector*) in the De sectore et radio. it was such a lamentable one. a position that he retained until his death. Broad Street. William Oughtred. John Aubrey. it fell to his lot to preach the Good Friday sermon. A competent mathematician.*30 Some evidence for the connection between. navigation.. Gunter had a gift for devising instruments which simplified calculations in astronomy. where he took his BA in 1603 and MA in 1606. At the end of this work Gunter wrote that'.. Henry Briggs. from London. On 6 March 1619 Gunter was elected the third professor of astronomy at Gresham College in succession to Thomas Williams.28 Gunter died suddenly in his rooms at Gresham College on 10 December 1626 at the relatively early age of 45. his tenure of office at Gresham overlapping that of Briggs by some 15 months. In 1623 he published an account.60 THE M A K I N 6 OF L O G A R I T H M T A B L E S Edmund Gunter and the Canon triangulorum Edmund Gunter was born in Hertfordshire. and fell to resolving of tri- angles and doing many fine things. fell into the hands of the leading mathematicians of the day. of this book. in English. many copies [were] transcribed and dispersed more than sixteen years since. written in Latin. Peter the Poor. 'Do you call this reading of Geometry? This is showing of tricks. Said the grave knight. man!* and so dismissed him with scorn. but they said that 'twas said ot him then in the University that our Saviour never suffered so much since his passion as in that sermon. The description and use of the sector. apparently gaining for Gunter 'the friendship of the Earl of Bridgewater. Georges. He was educated first at Westminster School and then at Christ Church. Southwark... the inveterate seventeenth-century gossip. He was buried in an unmarked grave in the church of St. recounts that:27 When [Gunter] was a student at Christ Church. and others.*29 It is probable that copies of this earlier version. of Welsh stock. and took his BD later the same year. in 1531. Gunter was not renowned for his preaching. (being of Oxford University). Later that year Sir Henry Savile was considering whom to appoint as the first Savilian pro- fessor of geometry at Oxford.. According to Aubrey he first sent for Mr Gunter.: so he came and brought with him his sector and quadrant. Oxford. In 1615 he became rector of St. and surveying. which some old divines that I knew did hear. which was in Latin...

THE M A K I N G OF L O G A R I T H M TABLES 61 Gunter and Bridgewater is supplied by Oughtred:31 In the spring. and the Sunne..' And not long after he delivered to Master Briggs.1 being at London. in Wright's translation.the first Earl of Bridgewater. mine own instrument printed off from one cut in brass. He was baron of the excheq- uer of Chester from 1599 until 1605... and questioned about the pro- jecture and use thereof. 1618. In the early 1600s John Egerton. . a work that did for sines and tangents what Briggs had done for nat- ural numbers..0000 & ad scrupulum prima quadrantis... another 'Oxford man'. He held many high public offices until his death in 1649. In addition. It may be presumed that it was the close relationship between Briggs and Gunter at Gresharn College which led to Gunter's Canon triangulorum of 1620.) The departure here from the convention that Briggs had already adopted of tab- ulating angular functions hi steps of one hundredth of a degree is striking. was a public fig- ure whose star was very much in the ascendant.. that is tables of artificial sines and tangents to a radius of 100000000 for each minute of the first quadrant. He viewed it very heedfully. (Trigonometrical canon. it is defined as follows:32 As for use and exercise sake. He received an honorary MA from Oxford in 1605 and became the first Earl of Bridgewater in 1617. Napier uses the same triangle in Chapter 5 of Book II of the Descriptio where. Gunter uses here the integer 100 000 000 for the whole sine. The title-page gives the foil title of Gunter's work: Cation triangufarttm sive tabulce sinuutn et tangentium artiftcialium ad radium 10000. after all. I showed him my horizontal instrument. often saying these words. 'It is a very good one. who then brought me acquainted with Master Gunter. let there be a spherical triangle not qoadrantal described on the superfices of the Primum Mobile PZS representing the pole. the title-page carries an engraving of a spherical triangle SPZ indicating clearly the proposed use of the tables in the solution of spherical triangles. the 'decimal point' in Gunter's title serves merely to guide the eye. the zenith. and was made a knight of the Bath on James Is arrival in England in 1603. with whom falling into speech about his quadrant. ten times greater than that used by Napier. much as we might use commas. went to see my honoured friend Master Henry Briggs at Gresham College. which afterwards I understood he presented to the Right Honourable the Earl of Bridgewater. to be sent to me. In the climate of the time perhaps Gunter felt that he could do worse than secure the patronage of such an important person as Egerton who was.

and which had been expounded for several years by Briggs at Gresham College. semi-quadrantally arranged. previously used by Napier. we may infer that the Canon was aimed at those who were already familiar with the solution of spherical triangles by the methods described in the Descriptio. but it is not impossible that it is the result of a desire to economize . 2.5 The title-page of Gunter's Canon triangulorum. This may have been a deliberate device to render the page more read- able. who was also to print Gunter's De sectore el radio in 1623. Earl ofBridgewater with whom. The Canon was printed by William Jones. is reproduced without explanation and it seems likely that the Canon was aimed at those already familiar with the solution of such triangles by the methods described in the Descriptio. The spherical triangle SP2. as we have seen. These tables are to eight figures. presents the logarithms of sines and tangents for every minute of the quadrant. published in 1620 and printed by William Jones. Since this triangle. in 1624. The Canon was the first table to be printed of logarithms of trigonometric functions to base 10. split into two groups of four by a vertical bar. previously used by Napier. The Canon triangulorum is dedicated to John Egerton. Gunter's table. like those of Napier. and Briggs*s great work.62 THE MAKIN6 OF LOGARITHM TABLES Fig. the Arithmetica logarithmica. Gunter seems to have had previous dealings. and which had been expounded for several years by Briggs at Gresham College. is reproduced on the title- page of the Canon triangulorum without explanation. a noted printer of mathemat- ical material. Leading digits are omitted where these do not change from the line above.

Thus. He does not refer explicitly to 'logarithms' but rather to the term Napier used in the Constructio.8317. and 32 minutes in the sixth column. Our table is useful for the solution of spherical triangles. And thus the second column and the fourth contain the Sines and Tangents of the angles in the first column. As before. all the numbers are integers. THE M A K I N G OF L O G A R I T H M T A B L E S 63 on the use of type. 'artificial numbers': But our Sines are not half of the chords.9134. when twice the radius is 20000 [ 0000 Take away from it the cosine of 51° 32' 979318317 The secant of 51° 32' is the result 1020611682 The rules used here are the usual trigonometrical identities.33 First. you should find sine 51° in the bottom part of the page. of which the following is a translation: To the reader devoted to practical learning.0) and sec(0) = l/cos(0). using as an example the latitude of London: So. Lord Merchistoe. At the end of the Canon is a letter to the reader. There then follows some explanatory material on how to use the tables.34 Similarly. It is of considerable interest and is transl- ated here in full. and die cotangent 9900. and the third column and the fifth. which for the sake of brevity we call the cosine. the common angle gives 9893. cos(0) = sin(90 .The first of these is of the angles from 0° to 45° degrees. but in their place their own numbers are substituted. For we avoid their multiplication . which we call 'artificial' for that reason. The remaining four in between contain the Sines and Tangents of these angles. they can be supplied by subtracting the cosine from double the radius. but it is a little easier to use. taking only 180 words or so in the original Latin. Gunter then distinguishes between the sines and tangents as usually defined and those in the table. Gunter's own description is very brief. with their first discoverer. the points and vertical bars being inserted only to guide the eye. of the angles from 45° to the end of the quadrant.7452 for the desired sine.0). the Sines and Tangents of the angles in the sixth column. When your calculation demands secants. which in London is 51° 32'. to which corres- ponds on the same line the sine of the complement 9793.0865. the tangent of 51° 32' is 10099. many greetings. the sixth. and the Tangents are not perpendicular to the extremity of the diameter. the general layout of the tables is described: Our table has six columns. divided from each other by two lines. the same as the tables of natural sines and tangents published by others. cot(0) = tan(90 . if you are looking for the artificial sine of our latitude.

and their division by subtraction. Regiomontanus seems only to have published a table of tangents. and that in 1504. Leading digits are omitted where these do not change from the line above. easier to use than the tables of natural sines and tangents published by others. and we use die same way of •working. expect more from us of this kind.35 It is at least possible that Wingate (or Ward in quoting him) made an error here. John Ward. His tables are. semi-quadrantatiy arranged. This table is to eight figures. . will perform that by addi- tion and subtraction only. This table. those in the Logarithmorum chilias primal.'. 2. Here we have explicitly stated Gunter's view of the utility of his tables in the solution of spherical triangles.64 THE MAKIN6 OF LOGARITHM TABLES Fig. have bound with ours [i. quoting Edmund Wingate. For we rest upon that foundation. he says. This may have been a deliberate device to render the page more readable. but it is not impossible that it is the result of a desire to economize on the use of type.6 A page from Gunter's Canon triangutorum containing the latitude of London (51° 32')...e. like Napier's. split into two groups of four by a vertical bar..e. And there is no need of more rules or examples. If you would like die same thing for rectilinear triangles. Gunter's tables in the Canon trianguhrum] the Logarithms of our friend and colleague Henry Briggs [i. by addition. presents the logarithms of sines and tangents for every minute of the quadrant. Farewell. states that Gunter's Canon'.3* It seems more probable that Wingate meant to refer to the Opus Palatinum of Rheticus that was . and if this work proves welcome to you. which heretofore Regiomontanus tables did by multiplication and division.

published in 1624. proba- bly dating from shortly after the publication of his Canon triangulorum. . First differences are printed inter-linearly. to ten digits. Briggs extended his pre- vious table in the Logarithmorum chilias prima to include logarithms. and from Briggs's reference to his 'intimate friends'. which gives many worked examples and refers specifically to Briggs's logarithms (the Logarithmorum chilias prima) and'my canon of artificial sines and tangents'. in 1628. established the techniques and conventions that were to underpin the two great canons of the next decade. In the years immediately following the publication of the Descriptio. There is some evidence. It was Briggs's intention to fill the gap in these tables but he was over- taken byVlacq who published the completed canon. Gunter's Canon triangulorum and Briggs's Logarithmorum chilias prima. it is likely that he and Briggs had a joint purpose in educating the mathematical community in the advant- ages of common logarithms and worked together for its fulfilment.37 The content of this manuscript indicates that it formed the basis of the second edition of the Canon triangulorum which was published in 1624. of the integers 1 to 20000 and 90000 to 100000. THE M A K I N G OF L O G A R I T H M T A B L E S 65 published in 1596. many of which appear in the British Library manuscript. the Arithmetica Iogarithmica and the Trigonometria Britannica. In his Arithmetica Iogarithmica. to base 10. This included a table of Briggs's logarithms extended up to 10000 and gave a rule for extending these up to 100000. again to fourteen places. There is in the British Library Manuscript Collection a manuscript by Gunter. that Gunter did subsequently come to believe that 'more rules or examples* were indeed necessary. In an extended preface Gunter described the general use of the table illustrated by worked exam- ples. however. That Gunter felt that there was no need for more rules or examples indicates that his Canon was intended for a readership that was already familiar with Napier's work. these two works. Briggs's Arithmetica iogarithmica and Trigonometria Britannica From Gunter's reference to Briggs in his Letter to the reader. or the version of those tables published by Pitiscus in 1613. In this table the comma is used both to separate the characteristic from the mantissa and to separate groups of five digits for ease of reading.

Briggs*s successor as professor of astronomy at Gresham College. as the Jrigonometria Britannica. should be calculated.. It was not until the end of the eighteenth century that any Eirther calculation of logarithms took place. the Opus Palatinutn of \ 596. But he hathe cut of 4 of my figures throughout. who hath done all the whole hundred chih'ades and printed them. together with log sines (to fourteen places) and tangents (to ten places). but the corresponding arguments in minutes and seconds of arc are also given. formed the basis of all logarithm tables published for the next 200 years or so. Briggs did not live to publish these tables: they were published posthumously in 1633. using the centesimal division of the quadrant.66 THE M A K I N 6 OF L O G A R I T H M T A B L E S In the same year Briggs wrote 'To his very good and much respected frende Mr. the intervals so expressed being 36 seconds. and by agreement the business was conveniently parted amongst us. with interlin- ear differences for all the functions. As described by Ivor Grattan- Guinness in Chapter 4. But this is typical of what we know of Briggs whose attitude he himself summed up in the words with which he closed his preface to Wright's translation of the Descriptio:'l ever rest a lover of all them that love the Mathematickes. and by some frendes whome my rules had sufficiently informed. . Briggs also calculated tables of natural sines (to fifteen places) and tangents and secants (to ten places). and its derivatives. but in view of their evident dose relationship it seems highly probable that Gun ter was one of them. at intervals of a hundredth of a degree..' The identity of Briggs's 'frendes' who were helping with the calculations can only be guessed at. in Cambridge":38 My desire was to have those Chiliades [thousands] that are wantinge betwixt 20 and 90 calculated and printed. and I have done them all almost by myself. That Briggs tabulated his sines every hundredth of a degree was in itself an.39 Briggs's two great works.. Vlacq's tabular layout differed slightly from Briggs s 1624 edition in that this time the first differences occupy a separate column.. The production of these tables. an Hollander. The division of the degree is thus centes- imal. Briggs. There is no hint in Briggs'$ letter of any irritation on his part against Vlacq for having anticipated him in the completion of a work that he. John Pell atTrinitie Coll. had all been sexagesimally arranged. in 1785 the French government decided that new tables of angular functions. but I am eased of that charge and care by one Adrian Vlacque. who also provided some additional material. the Arithmetica logarithmica and the Tngonotnetria Britannica. by Henry Gellibrand. had so nearly completed. and their logarithms. innovation: the great canon of Rheticus.

Briggs's table is of the logarithms to base 10 of the integers I to 20000 and 90000 to 100000. The comma is used both to separate the characteristic from the mantissa and to separate groups of five digits for ease of reading. . to fourteen places. THE MAKING OF LOGARITHM TABLES 67 Fig. First differences are printed intertineariy. 2.7 A page from Briggs's Arithmetic*! togarithmica. published in 1624 and printed by William Jones.

8 A page from Briggs's Trigonometrica Brltannica. published posthumously in 1633 by Henry Geltibrand. The equivalent angle in minutes and seconds is shown in the right-most column. This table is of natural sines (to fifteen places) and tangents and secants (to ten places). 2. . Briggs's successor as professor of astronomy at Sresham College. There are interlineal differences for all the functions. but divides each degree centesimally. together with log sines (to fourteen places) and tangents (to ten places). Briggs retains the division of the quadrant into 90 degrees. 44-GRAC>- Fig.

relatively expensive and written in Latin. Within two years of the publication of the Descriptio an English translation was published by Edward Wright. These two works were large. and the more publique good.40 Several attempts were made to publish these tables. There was a need for smaller and cheaper volumes.. a London teacher of mathematics. One of the earliest writers who sought to fill this gap was John Newton (1621-1678). al of them in English. During the Interregnum. a book of tables derived directly from those in the Descriptio. published his New hgarithmes. a staunch Royalist.44 The omission was rectified that year when Speidell published his A breefe treatise of sphcericall triangles. on arithmetic. This translation had Napier's blessing for in his preface to this edition he wrote that:43 I thought good heretofore to set forth in Latine for the publique use of Mathematicians [the Descriptio]. was entrusted to Gaspard de Prony with express instructions 'not only to compose tables which leave nothing to be desired in terms of exactitude but to make the greatest and most imposing monu- ment to calculation which shall ever be executed or conceived'. The popularization of logarithms Almost from the beginning. were not at once readily accessible to the student and practitioner. and were. In 1619 John Speidell. Newton. destined more for the coffee table than the working library. THE M A K I N G O F L O G A R I T H M T A B L E S 69 the Tables du cadastre.*5 But logarithms to base 10. none of them successful. despite containing no explanation of the tables nor any worked examples. comprising some 28 examples together with the tables which he had first published in his New hgarithmes. the utility of logarithms demanded that details of their use be published in the vernacular as well as in Latin—the then usual language of scholarship. A few attempts at recalculating logarithm tables were made in the nineteenth cen- tury. trigonometry. geometry. This work. taught mathematics in Oxford and in the years 1654 to 1660 wrote eight books. and . those by Charles Babbage (1827)41 and Edward Sang (I871)42 being the most notable. as set forth by Briggs in the Arithmetical logarith- mica said Trigonometma Britatmica. had gone through nine editions by 1627. perhaps. procured a most learned Mathematician to translate the same into our vulgar English tongue. writ- ten in English.. But now some of our Countrymen in this Island well affected to these studies.

The first 36 pages of the book. With his first book.. be as it were at once.. Newton's second volume. of Ross.47 which is usually found bound with the Institutio matlmnatica. but by being acquainted with many Books) might in a due method and a perspicuous manner. contain Oughtred's Trigonometry. The year 1657 saw the publication of Oughtred's Trigonometria. an important printer of tables in the third quarter of the seventeenth century who was. to become Hydrographer to Charles II. all to six places. It is one of the earliest works to adopt a condensed symbolism so that equations involving trigonometric functions could easily be taken in by the eye. Five of these volumes have a significant tabular content: tables of logarithms of the natural numbers and of angular functions and their log- arithms. explicitly derived from those of Briggs. the Institutio mathematics of 1654. . derived from those of Briggs and printed by Joseph Moxon. the Tabulce mathematics (1654). In its preface he wrote46 We. but his work in this period was devoted to the education of children. the Astronomia Britannica (1657)48 where he wrote It is therefore our chief and principal aim to shew how much of trouble may be avoyded in computing the motions of the heavenly bodies. printed by R.. the use of the decimal division of the degree. nothing that is our own. and W. contains tables of angular func- tions and their logarithms. together with a table of the logarithms of the first 10000 natural numbers. in 1662.49 a vol- ume of about 290 pages of which no less than 240 comprise trigonomet- ric and logarithm tables.. Newton made no claim to originality. and to encourage.. if only the form of our Tables were changed from Sexagenary into Decimal. all of them... presented to thy view. and later rector. a highly condensed work treating of both plane and spherical triangles. This only we have endeavoured.70 THE M A K I N 6 OF L O G A R I T H M T A B L E S astronomy. After the Restoration Newton became vicar.. Nowhere is Newton's espousal of the decimal division of a degree more clearly shown than in his statement in the preface to his third work. Herefordshire. that the first princi- ples and foundations of these studies (which until now were not to be known. Leybourn. He continued to be a prolific author. following Briggs. Newton's aim was twofold: to provide mathematical books in English. acknowledge that in this [book] we have presented thee with nothing new...

the number of Figures in the Tables falling short of that required and used in the rules. a Fellow of King's College. THE M A K I N G O F L O G A R I T H M T A B L E S 71 On 6 February 1655 Richard Stokes. particularly those involving the use of half-angle formulae. for I intend (since you were pleased to give your assent) to endeavour to print it with Mr. In these regions the rates of change of the tabulated functions are large and highly non-linear. It is clear that it was Oughtred's intention that the tables of logarithms of sines and tangents should consist of seven figures after the decimal point: in fact only six were printed. being 'much of their opinion who think translations the greatest enemy of Learning* although . and a second edition of the book. when finished I will send to you. Briggs his tables. Stokes felt compelled to apologize for writing in English. . Sir. in the appendix. The reverend Author has both discovered and amended the error. that is. wrote to Oughtred as follows:50 . for which volume the number of Figures was resolved on. . nor will you finde the least alteration [sic] except in the Rules for correcting the Canon which was occasioned by a mistake. Because of this error the rules given in the appendix for using the tables would not work. and upon the changing the Volume forgot to be altered. as you may there perceive. was rushed out within months of the first. Stokes wrote: The Book and Author are still the same. as farre as could be. In the preface to this edition. but something had gone badly wrong in the process of its production. sprung from the intention of Printing it in octavo. Cambridge.. The tables are clearly an integral part of the Trigonometria. near 0° for the logarithm of sines and tangents.. can be easily solved with the aid of logarithms and he gives a number of examples. Oughtred comments that trigonometrical equations. this time in English. Another passage in the preface sheds an interesting light on the attitude of publishers to the publication of scientific books in the vernacular. Following the tables there is an appendix of some ten pages which gives rules of interpolation in the trigonometrical tables for angles where linear interpolation is inappropriate. and near 90° for the logarithm of tangents. Undoubtedly the typesetting of the tabular portion of the Trigonometria was Moxon's piece de resistance. to know if it be according to your mind. I have procured your Trigonometry to be written over in a fair hand which.

. and Gunter in the Dictionary of national biography and the Dictionary of scientific biography provide useful information. logarithms had become ubiquitous among the mathematical community: tables were easily and cheaply available and many texts describing their use had been printed. a group centred on Gresham College with Henry Briggs as its driving force. version. Further reading The articles on Napier..51 By the end of 1657 Pell had devised a method now known as the Lagrange three-point interpolation formula.. Briggs. hyperbolic. the new appendix was essentially a translation of the previous. Within just a few decades of their invention.72 THE M A K I N 6 OF L O G A R I T H M T A B L E S he did not think 'that censure reaches things of this nature which are some- times of use to persons who it cannot be expected should understand any Language other than what their Nurses taught them*. Detailed references to the . namely of Collaterals. although different numerical examples were given. The starting point for the history of logarithms must be Charles Button's Mathematical tables. containing the common. The new appendix to the English edition begins with the statement: The Canon of Logarithtnes of Sines and Tangents was by the Author intended to consist of seven Figures after the Indices: but in this edition the seventh fig- ure was unhappily left out. to which is prefixed. a large and original history of the discoveries and writings relat- ing to those subjects. Although Stokes states that Oughtred was the one who discovered the error it is certain that John Pell (1611—1685) knew about it soon after the first edition was published. Being a descriptive catalogue of mathematical tables. and first and second differences.. This work went through eleven editions between 1785 and 1894. With this addition. Pell's papers contain many pages of numerical experimentation with methods of interpolation for use with Moxon's six- figure tables. What is striking is that their development from Napier's initial formulation was the work not of a few isolated individuals but rather the result of an almost explicit collaborative effort by the members of a closely connected group... A more recent treatment of the subject can be found hi James Henderson's Bibliotheca tabulamm tnathematicarum. 1926. so that now the Rules following will not hold in some of those tables.. and logistic logarithms. Latin. London. But they may hold indifferently wel. if unto the numbers in the Canon you joyne a circle or nul.

Longmans Green. llth edn. whose discussions on the invention of logarithms with John Napier are described by Wood (Athence Oxoniensis (ed. Glaisher. Oxford. 1617)\ Annals of the History of Computing. 3640°. This is Napier's value. Napier. A description of the admirable table of logarithms (trans. at 74 years of age* (quoted in Ward (note 23). implying that Briggs was born about 1556. 'Algebra in Napier's day and alleged prior invention of logarithms' in Napier tercentenary memorial volume (ed. M. Rabdology. London. be found in David W. There is some doubt about the year of Briggs"s birth. 1817. C. Knott). London. J. W. Baron. Joseph Mede of Christ's College. tr. It was in the Descriptio that Napier. Rider. 491—2) whose account provides some corroboration for Kepler's assertions. De thiende. vol. New York. 1889. following the invention of Stevin.'The first calculating machine (John Napier. London. 1974. Hawkins describes in detail the construction and use of the promptuary and gives an almost com- plete translation of the appendix. pp. 90-6. P. 7. 8. 2. In Art 5 he wrote that 'in numbers distinguished. 9. 1585. and their contemporaries can.. whatever is written after the period is a fraction. cols. Notes 1. life. Edinburgh. 10 (1989). L. R. Stevin. London. intro. 1915. mathematical'. 1834. Quoted in: E Cajori.. 609-13. 104. Mass. 4. William Frank Richardson. THE M A K I N G O F L O G A R I T H M T A B L E S 73 work of Briggs.. 6. The construction of the wonderful canon of logarithms (trans. The Halifax parish registers give the baptism of'Henricus filius Thome Bridge de . Margaret E. wrote on 6 February 1630 that 'Mr Henry Brigges of Oxford. pp. Napier. Cambridge. W. is lately dead. the great Mathematician. J. p. E Hawkins. Cambridge.. 120). This is Hawkins's English translation of the original Latin. W. Waters' The art of navigation in Elizabethan and early Stuart times.} argues that this 'certain Scotchman* was Dr John Craig.'Napier. I have relied on the DSJB article for biographical information on Napier. 'Tables.. Blackwood. vol. his lineage. Gillespie). with a history of the invention of logarithms. in Encyclopaedia Britannica.'. Bliss). 1910-11. used the decimal point as a separator of units and tenths. by a period in their midst. 26. 1616. but log 10 is more nearly 23 025 850. 5. Robin E. p. Napier. See J. G. 9 (ed. Napier (Memoirs of John Napier of Merchiston. 3. Dictionary of Scientific Biography. 1965. 39. 1990. 1958. Leyden. and times. Gunter. John'. A detailed discussion of the Gresham College circle can he found in Christopher Hill's Intellectual origins of the English revolution. See: S. Edinburgh and London. J. Wright). E. C. pp.93. C. Macdonald).

(l). the angle whose sine is 1/10...'Napier's logarithms and the change to Briggs's logarithms'. 26. 1915. in Napier tercentenary memorial volume (ed. 194-5. 16. G. 14. W.)). a date which may be misleading in that then. 10.54.74 THE M A K I N 6 OF L O G A R I T H M TABLES Warley' OH 23 February 1561 (new style). No previous translation of this preface has been published..York.e.lO. G. Gibson. 13-25. 18. baptism did not always occur within a few days or weeks of birth. p. London.(2. That is. Knott). Blundeville. London. though.Adamson. 11. Sloane 917. The contributions of Briggs and Ms contemporaries to the science of naviga- tion are treated extensively by Waters (D. E. is also extensively annotated. Waters. This quotation is from Gibson's translation of Briggs'$ introduction to the Arithmetica hgarithmica of 1624. that the identification of Henry Briggs with the Henricus Bridge of the parish registers is far from certain.lO. In die Latin languages canoncs was the name given to introductory instructions as to how the tables were to be used. 17.'The administration of Gresham College and its fluctuating fortunes as a scientific institution in the seventeenth century'. Knott). Sampson. . as now. in Napier tercentenary memorial volume (ed. is copiously annotated and contains extensive manuscript tables of differences.'Bibliography of books exhibited at the Napier tercentenary celebration. 13. A. The theoriques of the seven planets. vol. and this soon became its stan- dard meaning. 19. co. T. Certaine errors in navigation. There are two copies of this work in the British Library. Quoted in Ward (note 23) p. Brige. in the Manuscript Collection. Bridg(e). Brydge. R.. London. R.The Publications of the Yorkshire Parish Register Society.W. Brygg. 1610. Brygg(e).54. The second copy. 121. but by the beginning of the Middle Ages the word was applied to a complete set of tables. bound with Gunter's Canon triangulomm (C. 37. 1602.Wright. Sampson's in extenso quotation of the Latin original is incomplete. It may be remarked.e.1910. Wood's statement that at the time of his death in 1630 Briggs was 'aged 70 or more* is not wholly inconsistent with either date. pp. See The parish registers of Halifax. G. passim. Longmans Green. (ed. For a detailed account of the foundation and early years of Gresham CoEege see I. The family name Briggs is totally absent from the registers but there are many occurrences of its variants Brigge. E. by a different person. 1958). Brygghe and Bryghe. Brege. 1. C. and Henry was a common first name. 1915. C.A. July 1914'. London. That at shelf mark C. Bryge. and contains a number of worked examples. Crossley). 15. Longmans Green. The art of navigation in England in Elizabethan and early Stuart times. 12. History of Education 9 (1980). London.

London. 1632. 77. . Gray. 49. Oughtred. G. or The mathematicall ring.'Gunter. 1873. W. The term 'cosine' seems to have been Gunter's invention. f. in a pamphlet called Gmmtnclogia. Elder. British Library Manuscript Collection. Dictionary of National Biography. 37. or Minfica logarithmorum projectio tircularis. 122. J.Woodbridge. R. . a view shared by Glaisher (note 36). Glaisher. Brief lives (ed. which was published in 1617. The first example of the term 'Briggian* seems to have been in 1774 (Philosophical Tran- sactions. 59. C.Aubrey (note 27). Napier's Mirifid logarithmorwn canonis descriptio of 1614. I am grateful to Carol Bostock-Smith for her ready help with translations of both Briggs's and Gunter's (see note 19) Latin. 30. These are sometimes called Briggian or Briggsian logarithms. 34. a primis fundamentis accurate extructum: 21. 25 November 1892. (See note 18. A description of the admirable table of logarithms (trans. 41. 33. The assertion in the DSB (note 1) that it refers to the Rabdologiae. 32. THE M A K I N G OF L O G A R I T H M T A B L E S 75 20. must have been written in about 1607. It seems clear from the context that this is a reference to Napier's Constructio which appeared in 1619. 1618. 28. Gunter. p. ipsis Sinubus proportionates. The general use of the former canon and table of hgatithmes. 36. Smith. edited by his son. The just apologie of Wil: Oughtred. London. The variant 'Briggsiaii' seems to have been first used in 1892 (Science. p. W. 279. Napier.(2. 26. p. 33. 64. Report of the British Association on mathematical tables. 31.per oequationes Algebraicas. This Latin work.54. p. 350-1. p. London. 35. 1982. Barber).) 25. 23. Shane 910. London. p. is probably mistaken.. E. 29. L. 24. 307). J. 1740. p. vol. against the slaunderous insimulations of Richard Delamain.lO. 1890. pp.J. and in any case. The date of 1603 given in the DNB is unlikely to be correct: Gunter's statement is unambiguous. J. Jo the English gentrie. The lives of the professors of Gresham College. in 1603 he would still have been an undergraduate. p. the New projection of the sphere. E. London. . The terms 'characteristic' and 'mantissa' seem to have been coined by Briggs and first appear in his Arithmetic logarithmica of 1624. J. 27. Ward..e. Edmund'. pro singulis Gradibus & graduni eentisiniis. & differentias. Quoted in: J. Baydell. and all others studious of the mathematicks .Aubrey. 22. Wright). Ward (note 23) p. '. 121. 23.). 223).

. 1627. and reduced into Theorems. J. and theory of the planets decimally by trigonometry. 43. London. or hyperbolic. My translation of the original French: '. Speidell's table of such logarithms. Tabula. corrected..e. Tables of the natural Sines. Institutio mathematical.000. Dialling and Navigation. so that by addition only. Tangents and Secants. JHgonontetria Britannica. Astronomia Britannica. Quoted in:J. lately published by Doctor Ward.000. Glaisher (note 36). C. Speidell. J.. Tangents. Glaisher. and their calculators'. by the Honourable Lo: John Napair Baron of Marckiston. incorrectly. according to the Copernican systeme as it is illustrated by Bullualdus. Shane 3498. niais a en faire le monument de calcul le plus vaste et le plus imposant qui cut jamais etc execute ou nieme concu. stated that the table in this work was the first published example of what are now called Naperiao. L. 44. was. p. Newton. As also. 42.. A new table of seven-place logarithms of all numbers from 20. 4th Ser. and the Logarithms of the Sines and Tangents to every degree and hundred part of a degree in the Quadrant. 46. 1827. A Mathematical Institution Shaving the Construction and Use of the Naturall and Artificial Sines.. 41.. J. J... Briggs. require not at all any skill in Algebra or Cossike numbers. exhibiting the doctrine of the sphere. non seulement a composer des Tables qui ne laissassent rien a desirer quaot a Inexactitude. the twelve cases of an oblique sphaericall triangle.. 1619. logarithms to base e). and then of natural numbers rather than trigonometric functions.. H..000.. . excudebat Petnts Rammasenius. Napier (note 6).000 to 200. 39. in Decimal Numbers and also of the Table of Logarithms. Sang.000. and the easie way of calculation. 1657.. E. London. Fitted for the meridian ofLondon. New logarithms.76 THE M A K I N 6 OF L O G A R I T H M T A B L E S 38. Gouda.. and Secants. Babbage. These being extracted from and out of them. London. logarithms (i.. 1654. mathematical.45 (1893). London. and by tables. Speidell... 'On early logarithmic tables. A breefe treatise ofsphericall triangles. etc. Philosophical Magazine. they may be effected. J.. 45... 1633.. they being fast over seene. unth Their more particular application in Astronomic. being all extracted out of one Diagram. J. 376-82. 48. Their common Radius being 10. u'herein is handled the sixteen cases of a right angled triangle. Newton. and amended.. with the total sine in the first place. Newton. London. . London. London. With a Table of Logarithms of all absolute numbers increasing by naturall succession from an Unite to 10.. The first invention whereof. 1871.. In the general solution of ant Triangle whether Plain or Spherical.*. 56. did not in fact appear until the second edition of New logarithms in 1620. 1654. quoted in. 47. Tables of logarithms of the natural numbers from 1 to 108000.W. It is often. British Library Manuscript Collection. 40. Preface.

50. W. was published in the same year. British Library Manuscript Collection. Trigonometry. The English edition. p.J. Rigaud). ff. 51. 1841. THE M A K I N G O F L O G A R I T H M T A B L E S 77 49. Oxford. 82. vol. Oughtred. London. Add MS 4415.. 1657. Trigottometria. S. t26v-131. . Correspondence of scientific men of the seventeenth century. I (ed.

perCent. TMe Numbers in this Breuiat ranft alfo be dice- med NumeratorSjeachof them hailing for Denomina- tor IOOOOOOO. Now foUovr the Qgcftioiis i In working whereof.we will vie this Brcuiar. The .I* The Breuiat of the Table of\Q.

.1 Table of compound interest from Richard Witt's Arithmetical questions. 3. starting with the relatively simple concepts of compound interest and Fig. 1613. The actuarial profession was formed in 1848 but long before then mathematicians were producing tables in these two areas. 3 History of actuarial tables CHRISTOPHER LEWIN AND MARGARET DE VALOIS Actuaries make financial sense of the future by combining techniques of risk and finance.

this remained fixed despite the death of some subscribers. Klingenberg published a life table which showed the expected number of survivors for every year in future. Life tables show the number of survivors in a population which is reduced with the passage of time by one event. In 1653 Poul Klingenberg (1615—1690) moved to Copenhagen and persuaded the King to establish the world's first tontine (an idea which had been suggested by Lorenzo Tonti. such as the resignations. In order to market the scheme. Once the State's total annual payout was established. Compound interest tables A surviving fifteenth-century Italian manuscript includes compound inter- est tables showing how much an invested sum will accumulate to after various numbers of years. In 1558. in an arithmetic textbook published at Lyons.3 However. for use in banking transactions. for different purposes.4 Only a few tables were given. Compound interest was in use among ancient civilisations. so each of the surviving subscribers received a sum which gradually increased each year.80 H I S T O R Y OF A C T U A R I A L T A B L E S probabilities of survival. it was not until the second half of the sixteenth century that any serious attempt was made to publish compound interest tables. Other tables were developed later.2 The subscribers to a tontine paid the State a capital sum in order to receive annual payments for the rest of their lives. 250 for 40 years and 35 for 60 years. deaths. The table suggested that of 2000 children aged up to eight. It is believed that this was not based on scientific data but was purely illustrative as an advertising aid. namely death. an Italian banker who lived in Paris). This chapter shows how actuarial tables have devel- oped over the past four centuries. These included tables calculated at 4% interest per .1 though as far as we know they did not have any tables relating to it. only 900 would survive for 20 years. It was probably not until the Middle Ages that compound interest tables first came into being. Jean Trenchant included a table on simple and compound interest. but around 1900 the con- cept was extended to cover multiple events. These life tables were then combined with compound interest tables to facilitate evaluation of the sums of money to be paid for life assurance and annuity contracts. Proper life tables based on real data started to emerge later in the seven- teenth century and became a fundamental tool. and retirements in a workforce.

16. There was also a table showing the total accumu- lated result of investing a series of payments of the same amount each period. The tables were 'the other way round' from those which had been published by Trenchant 27 years previously. because they avoided the need for division. 17. In other words. i. if 100 units of currency is invested. He then adds to this the present values corresponding to payments of 500 in 2 and 3 years' time respectively. There was also a set of tables calculated for interest rates from 1% to 16% and also for rates of interest corresponding to 15. This is called the 'present value' equivalent to 500 in a year's time. showing the accumulation over a number of periods of an initial investment of one unit of currency. In 1585 a more detailed work on compound interest was published by the famous mathematician. rather than the sum to which a sum invested now would accumul- ate by the end of the period. etc.04 raised to successive powers. because. a table based on interest at 10% per year showed the accumulation of an invested sum for complete months and also for periods of years up to 6. as part of his own arithmetic textbook. The latter is shown to be marginally worse. 3. invested at the end of that period. and 22 years* purchase of freehold properties. they showed the present values of a sum due to be received at the end of each period. Stevin's table showed 1 divided by 1. Trenchant discusses whether it is better to receive 4% interest per quarter on a loan or regular payments of 5% per quarter for 41 quarters with no return of the sum.5 A number of worked examples were included for problems of both simple and compound interest. He shows how to use the 4% table in reverse to find the sum which must be invested now in order to have 500 at the end of the first year. 2. He quotes a case where the rent of a farm is 500 per year (in arrears) for 3 years and it is desired to pay it in advance by a single payment. . For all problems of the type presented in Trenchant's book.e. the extra unit from the annuity. H I S T O R Y OF A C T U A R I A L T A B L E S 81 period. whereas Trenchant's table showed 1. 19. each quarter for 41 quarters. a table of 1.04 raised successively to the power of 1. 18. Trenchant also points out that to receive 4% interest per quarter is better than 16% per year. i. As well as these tables at 4% interest. Simon Stevin of Bruges. This is called an annuity certain.04 raised to successive powers.e. Stevin's tables were in a more convenient form. accumulates at 4% per quarter to slightly less than the 100 of principal which would be returned at the end of that time if the loan were selected instead. 21. Stevin also provided tables corresponding to the total present value of a series of payments to be received at the end of each period.

1 at the opening of this chapter).04 =0. invested now will accumulate to after 1. it was becoming generally accepted that charging interest was in order. i.. The book begins with a table (or 'breviat') calculated at a rate of interest of 10% per year (see Fig.043 = 1.04 1/1. This table is used in various worked .042 = 1.043 = 0.1 raised to the power of the number of years concerned.years. However. In an example where someone pays a lump sum instead of an annual payment for 22 years. However. calcul- ated at an interest rate of six and two thirds per cent. which allowed interest to be paid up to 10% per year. an accumulation table showing how much £\. and there is no evidence that risk entered into his thoughts at all. 2. by the present value of a payment of 1 at the end of that period. It delved deeply into the subject in a very practical way.0816 1/1. a table of 1.82 H I S T O R Y OF A C T U A R I A L TABLES Accumulation table and present value table for a single payment Period of years Sum to which Present value 1 invested now will of 1 at end of accumulate in period period 1 1. By about 1600. all Witt's payments were certainties. and books on compound interest started to appear. but instead demonstrates how these accumulated values can be derived from his present value tables for a series of payments by dividing the present value of an annuity of 1 per year for the number of years concerned. The principal compound interest book of this period6 was written by Richard Witt. and published in 1613. he does not give similar tables for other rates of interest. 1. Even after the Act became law. It is.042 = 0.96154 2 1. it was found to be ambiguous.92456 3 1. a 44-year old mathematical practitioner. to be received on definite dates. In England it had for a long time been illegal to charge interest on loans (though this did not prevent it happening underground!) until the passing of an Act of Parliament in 1571. he shows how to find the rate of interest underlying the transaction by inspection of the tables for various rates of interest. this is in the same form as the accumulation tables given by Trenchant. however. quite simply.e. .88900 Stevin also gives a table of the accumulation of a series of payments.. 3.1 2486 1/1. and it is evident from the clarity of expression and the care which was taken that the author thought in much the same way as modern actuaries.

However.61051 and multiply the result by 10 to give the correct result of 3. The original purpose was to warn of the rise of epidemic diseases. Perhaps the most interesting7 is the Table oj leasses and interest printed by William Jones dwelling in Red-cross street (London). constructed by a London draper. The book contained 124 interesting worked examples.61051.8 The Bills of Mortality were printed statements of the numbers of people who died each week. H I S T O R Y OF A C T U A R I A L T A B L E S 83 examples to obtain other compound interest functions as necessary. he goes on to give tables of these other compound interest func- tions.7908. One oweth . His method would be to subtract 1 from this to give 0.£900 to be paid all at the end of 2 years: he agreeth with his creditor to pay it in 5 years. though it may have been a Mr Aecroid. with a view to drawing conclusions about London's population and other matters.The author's name is not given.73 Life tables The first scientifically based life table was. per annum interest. suppose that you wish to find the present value of a series of pay- ments of £. every year a like sum. His conclusions were published in a pioneering book in 1662. They demand what each of these 5 payments shall be. decided to investigate the London Bills of Mortality for a long series of years. divide this by 1. Witts table shows that after 5 years 1 will have accumulated to 1. classified according to the apparent cause of death. reckoning 10 per cent. having demonstrated these methods. For example. and interest upon interest. to avoid the user having to make tedious calculations. The main point of interest lies in the fact that the book contains a description of the rather odd method of construction of the tables. so that the better off could flee in time to the perceived safety of the .61051. Numerous other writers followed Witt in producing compound interest tables. John Graunt. of which the following is an example: Q70. curiously enough. starting with a payment in one year's time. viz. His use of processes such as this shows that the mathematics of com- pound interest were fully understood. who had a shop in Birchin Lane. The book was clearly intended as a practical ready reckoner for everyday use by valuers and others who were concerned in the renewal of leases.1 per year for the next 5 years. such as the plague.

(The data were not subdivided by age. 36 die before the age of 6. the other figures in the table are purely speculative.) He also asserted that perhaps one of these 100 indi- viduals survives to age 76. Life table (based on Graunt) Age Number alive Deaths before next listed age 0 100 36 6 64 24 16 40 15 26 25 9 36 16 6 46 10 4 56 6 3 66 3 2 76 1 1 86 0 •— It is apparent that. he then sought 'six mean proportional numbers between 64. although the number surviving to age 6 (i. Each 'number alive* after age 6 is exactly or approximately equal to five-eighths of the . sworn to their office*) went to view the corpse and make enquiries. 64) is based on data from the Bills of Mortality. where the week's Bill for London as a whole was made up and printed on the Wednesday. They reported their findings to the local parish clerk and he. took a weekly return of the deaths and christenings to a cent- ral office. To get the numbers of survivors at intermediate ages at ten-yearly intervals. When anyone died. so Graunt had to guess which causes of death were most likely to relate to children under 6. Graunt estimated that of 100 children conceived.24 are assumed to die before age 16. of 100 people conceived. Graunt's table shows that. Upon analysing his data. and the one which survives 76'. leaving 40 surviving to age 16. the church bell was rung and then the official searchers (whom Graunt describes as 'ancient matrons. It was published on the Thursday and taken round to the families who were willing to subscribe four shillings per annum. and so on. every Tuesday night. to establish what the person had died from. the remainder living at 6 years.84 H I S T O R Y OF A C T U A R I A L TABLES countryside.The data underlying the Bills were collected by the parish clerks. Of these 64. leaving 64 surviving to age 6. 36 die before age 6.e. and are designed to secure a smooth progression.

and . Graunt's mathematical concept was very far from the truth.4%.2 A London Bill of Mortality. which is about 1 in 18. previous one. HISTORY OF ACTUARIAL TABLES 85 Fig. independent of age. 24-31 August 1680. He does not seem to have noticed the serious discrepancy between the overall annual mortality rate implied by his life table (if the population is assumed to remain the same size). However. 3. This is equivalent to an annual survival rate of about 95.

and made an impact throughout Europe. There has been much confusion about the meaning of the 1000 persons shown at age 1 'curt*.The important point about these data was that they were classified by age at death. in the table. His life table is shown in the accompanying illustration. or 1110 live births. pastor of Breslau. and sent to the Royal Society in London. In fact it means 1000 children living that will be age 1 at their next birthday This corresponds to about 1174 children conceived. The underlying data were collected by Caspar Neumann. .3 Halley's life table. from the church registers there. he might well have concluded that the obvious way to make the figures consistent would have been to assume for his life table a mortality rate for adults which in some way increased with age. 3. Halley's paper* was published in the widely-circulated Philosophical transactions of the Royal Society in January 1693. with many writers mistakenly assuming that it meant 1000 births. for the period 1687—91. We now come to Edmund Halley's life table. If Graunt had realised the existence of this discrepancy. unlike the London Bills of Mortality.10 Halley's life table later proved to be a sufficiently accurate estimate of mortality for many practical purposes. Although Halley had pointed out how to use the table in conjunction with compound interest functions in order to obtain the purchase price which should be paid for a Rg.86 HISTORY OF A C T U A R I A L TABLES the overall annual mortality rate of 1 in 27 which was shown by his data from the deaths in the Bills of Mortality and his estimates (which were probably reasonably correct) of the size of London's population. 1693. in which case the table would have been much more correct. which formed the basis for his great work on the valuation of life annuities—the creation of actuarial science.

some Continental countries had sold life annu- ities to the public for much of the seventeenth century. where the immigrants had an age distribution different from that of the native population. The investigators took the records of Government annuities which had been sold during a particular period and followed the annuities through over a period of many years. an equal number of them would die each year until age 85 inclusive. The records were kept accurately. Several Continental investigators decided. Unlike Britain. H I S T O R Y OF A C T U A R I A L T A B L E S 87 stream of regular payments throughout life. This led to a very simple calculation indeed. No information was available about the corresponding numbers of people alive at each age. Alternatively. " This method assumed that. and after a few years this enabled writers such as John Smart12 and Thomas Simpson13 to construct their own life tables from. However. so that none would be left alive at age 86. until all the annuitants had died. Because of such problems. leading to a population bulge at the ages to which those incomers had now attained. Such reasons might include immigration. the London Bills (which tables generally showed higher mortality than Halley's table—perhaps because of the widespread social evil of excessive consumption of gin at that time). the population age structure might be very far from stationary due to past fluctuations in the birth rate. for some reason. From 1728 onwards the London Bills of Mortality recorded deaths within broad age groups. in the case of a group of lives who were all the same age (over 12). because the payments depended upon them. Even if neither of these causes was applicable in a particular case. with an error of less than 5% at all ages. to try an entirely different approach. and distortions in the results . the calcula- tions were very laborious in practice.This could lead to severe distortions if. following which the city population would probably rapidly have been replenished by young adults coming in from the surrounding countryside.e. for example. there might have been a plague epidemic which carried off much of the popula- tion at all ages some years previously. therefore. Consequently. because there were no population censuses. The method did not work so well with later life tables. it was difficult to be confident of the numer- ical results obtained from the analysis of Bills of Mortality. i. the population had not yet reached a stationary state. the mathematician Abraham de Moivre invented an approximate method which enabled life annuities to be valued without using a life table. however. these life tables suffered from the fact that they were constructed only from the numbers of deaths at each age. a 'life annuity'.

88 HISTORY OF A C T U A R I A L TABLES obtained were therefore kept to a minimum.4 The Northampton table (by Richard Price).. 3. Fig. . In due course population censuses became available and it became possible to relate the deaths occurring in a year to an accurate figure for the number of people alive in a year at the same age. Mortality rates seem first to have been calculated in this way by Pehr Wargentin'7 in Sweden in 1766. Shewing the E X P E C T A T I O N S of Human Life at every Age. deduced from the Northampton Table of Obfervations. The mortality rates can then be converted into survival rates which are 38] T A B L E S . In later years it was seen to produce out-of-date figures. TABLE VII. This method became known as the 'cohort' method. because it followed through a cohort of lives until they had all died. which could not keep up with the significant improvements in mortality which started to emerge due to medical advances and improving sanitation in towns. Life tables of this kind were produced in the 1740s by Strayck14 and Kersseboom15 in Holland and by Deparcieux16 in France.

James Dodson. the last of whom died in 1783. too.18* Even that table proved too conservative and further surpluses arose. which were produced by Dr John Heysham. However. In 1829 John Finlaison. adopted the Northampton table. since it helped to ensure the establishment of the life assurance industry on firm financial foundations.This gave all the offices a built-in margin for adverse con- tingencies and profit. However. and perhaps showed somewhat too light mortality.The report demonstrated conclusively that the terms on which the . even for life assurance. many people felt that it was unsatisfactory to use a table which differed so much from the actual experience of the offices. and in 1782 adopted a new lower scale of premi- ums.20 the Government's Actuary (who was known as the Actuary of the National Debt) reported on a number of painstaking mortality investigations which he had carried out on a 'cohort' basis. Some of the resulting tables were produced separately for men and women. based on the Northampton life table produced by Richard Price. Although the table was based on relatively scanty data. This is the 'census' method usually employed today. which may not have been a bad thing in those early days. Other life companies were now starting up and they. which was founded in 1762 as the first life assurance company offering long-term business financed on proper scientific princ- iples. H I S T O R Y OF A C T U A R I A L T A B L E S 89 combined to build up an accurate and up-to-date life table. a London mathematician. and considerable interest was therefore awakened when in 1815 Joshua Milne'9 published a table showing lighter mortality. derived from the Bills of Mortality for the City of Carlisle between 1779 and 1787. the Society actually experienced lower death claims than had been expected. constructed his own life table as a basis for calculating the initial tables of premium rates for the Equitable Life Assurance Society. was poorly graduated from. the reasons for the relatively light mortality experi- enced by the offices may have included the fact that life assurance was mainly used by the better-off professional classes. The first showed the experience of 1002 nominees of the English tontine of 1693. no doubt through competitive pressures to quote lower premiums in order to attract new business. it was in widespread use by 1853. Apart from possible defects in the construction of the Northampton table. and that only people in good health were accepted. one age to another.18 Prior to its formation the Society had applied for a Government charter but this was refused on the ground that the premiums might prove to be too low. This was his famous Carlisle table. The other investigations related to the participants in other Government annuity schemes of the eighteenth cent- ury.

Halley had looked at each future payment of the annuity and worked out from the life table the probability of survival to that age. in other words the table needs to be 'graduated'. have other diseases risen? Is there an age beyond which humans cannot live. This probability was then multiplied by the present value (obtained from compound interest tables) of £\ paid at the time the annuity payment in question fell due. even if more individuals approach that age? Any table of mortality rates produced from crude data contains irregularit- ies due to chance fluctuations.The Government. climate.90 H I S T O R Y OF A C T U A R I A L TABLES Government was still selling annuities to the public were too generous. data errors. While this might not sound too difficult for just one payment.21 who laid out his calculations in columns with a summation backwards from the end of life. This enabled the value of a life annuity at any age to be derived quite simply by one multiplication and division. (For accurate results all the multiplications involved had to be done with the aid of tables of logarithms). etc. The method was then improved in 1785 by . Finlaison used a moving-average method for this purpose. value of the annuity. which until then had ignored Finlaison's warnings. it was very time consuming and liable to error for a whole series of future payments. and social class? Does mortality improve over time? Has male mortality changed at the same rate as female mortality? With the decline of smallpox due to vaccination. The first step was taken in 1772 by William Dale. It was realised that there would be considerable efficiencies if tables could be provided which con- tained the summations and some of the multiplications already done within them. the results of which had then to be summed to get the total. Finlaison's report also raised some fundamental philosophical ques- tions. Technical Improvements In life tables Towards the end of the eighteenth century new methods were starting to be adopted to ease the burden of calculation involved in an accurate application of Halley's method of valuing life annuities. now adopted new tables (which were used until 1884) and saved considerable sums. which were doubtless discussed in actuarial circles: Is there a law of mortality? Does mortality vary by district. These need to be smoothed out.

e. which were produced in 1824. they increasingly came to be calculated from mortality rates derived from the experience of a short period of recent years. One of the methods he proposed was for the correction of an isolated error in a run of figures. H I S T O R Y OF A C T U A R I A L T A B L E S 91 Tetens22 and eventually came to be known as the method of'commutation columns'. in line with the latest advances in medicine and sanitation. there was the practical consideration that the probability of death could then be found at any age from specified parameters. and the time consuming col- lection of data would be less necessary. The first well-known attempt to devise a law of mortality was due to Benjamin Gompertz. in 1839 Woolhouse27 assumed that new entrants joining part way through a year had half as much exposure as indi- viduals who were there at both the start and end of the year.24 Thomas Rowe Edmonds. Attention now started to be paid to the best method of determining how many people were in fact exposed to the risk of death. He considered this method to be superior to the previous method of using moving averages (such as that which had been used by Finlaison in 1829). the differences between the first differences) were also . formulae of this type are nevertheless still used occasionally today as an approximation for some practical purposes. Woolhouse28 was the first to explain the method of graduation using differences in 1866. He wanted to find simple methods of adjusting the crude rates so that they ran smoothly from age to age. In principle the mortality rate is the number of deaths at a given age in the period. Because of the emerging need for life tables to be up to date. Throughout the nineteenth century there was much interest in the search for a mathematical law of mortality. The sec- ond differences (i. As will be seen below. For example. and indeed the search for such a law has been discontin- ued.23 and his law was subsequently refined by Thomas Young. Later in the nineteenth century there were developments in the meth- ods used to graduate tables produced from crude data.26 Despite the fact that the Gompertz formula is no longer considered to be a universal law of mortality.25 and William Makeham. which still continues in use for some purposes. The initial step was to calculate the first differences between the successive crude figures. Apart from the philosophical interest which would result from the discovery of such a formula. it was partly a lack of a proper understanding of how to calculate exposures to risk that led to the relatively short life of the Highland Society of Scotland's sickness tables. divided by the number of people who were exposed to the risk of death at that age.

OWC Edmonds was the first to use Gompertz s formula to produce mortality tables. add together the two remainders and divide by six. In 1825 he formalized his law and checked it against the life tables of the time. He postulated that the mortality rate (or. In 1826 Thomas Young examined graphical curves of mortality to see whether they conformed to a formula.'to afford the means .. He produced a high-powered polyno- mial that would probably have resulted in spurious results if used in practice. x. will be the correction to be made. The formula proved particukrly useful in practice for calculating functions for multiple lives. with its proper algebraic sign.92 H I S T O R Y OF A C T U A R I A L T A B L E S The search for a taw of mortality Benjamin Gompertz outlined the first sketches of such a law in 1820. rather than just eliminate isolated errors. In 1859 William Makeham suggested that the Gompertz law was not accu- rate and started to develop an improved formula for the force of mortality: Bf + A where B. Woolhouse then went on to consider difference methods of adjustment to make a whole table proceed more smoothly. strictly speaking. Irregularities in the pattern of the differences would then iden- tify the crude figure which was in error. alongside graphical and other methods. was given by the formula a* where B mid c were constants. The result. Difference methods of graduation are still widely used today. hi 1832 Thomas R. c and A are constants. calculated. a very similar function called the 'force of mortality') at a given age. Life office mortality tables In 1838 a Committee was formed in order to conduct an investigation into the combined mortality experience of 17 life offices.. The correction to be made to that crude figure would be obtained by the following rule: From the middle second difference subtract separately the preceding and following differences.

At first a new committee was set up and new data were collected for each investigation. compared with 2248 expected by the table. seized the initiative and lepublished the main results3" with comments and monetary functions. Jones pointed out that the continued use of the Northampton table by the Equitable meant that it had only 1489 deaths in the 12 years preceding 1829. with lower mortality than men thereafter. The tables showed survivorship rates but these were not at that stage combined with compound interest factors to derive monetary functions which would make the tables easier to use. H I S T O R Y OF A C T U A R I A L T A B L E S 93 of determining the Law of Mortality which prevails among assured lives. However. Jenkin Jones. Jones also concluded that the Irish had higher mortality than the British. though generally giving a lower expectation of life than that table. Overall the new tables agreed very closely with the Carlisle table. the actuary to the National Mercantile Life Assurance Society. thus enabling a continuous mortality investigation to take place. Each new set of tables that was produced came to be known as an inves- tigation. The tables were not used very much in Britain but were widely used in mainland Europe. Details of nearly 84 000 policies were collected. The method was in theory less accurate than that which had been used in previ- ous investigations and proceeded by taking an annual census of the population exposed to risk and calculating the rate of mortality as a/b. and pio- neered a spirit of co-operation between the offices in this area which has lasted to the present day. with the need to prepare data retrospectively. The Carlisle table was better in this respect but it was so irregular from one age to another that it was unsuitable as a basis for calculating the premiums for temporary assur- ances. and that women had higher mortality than men between ages twenty to fifty. The old methods had involved considerable trouble and delay. even into the early twentieth century. The resulting life tables29 were published by the Committee in 1843. in 1912 Elderton and Fippard31 showed how the census method of mortality table construction could be applied to life office data. in return for access to the results.' Fifty-eight offices funded the project. where a was the number of deaths in the year and b was the mean population plus one-half of the deaths. Before the Committee had come to any conclusion over the calculation and publication of tables of monetary func- tions. Mortality tables based on life office data continued to be produced in later years. and it was felt that . Their main significance was that they promoted the use of life office mortality data rather than population data when setting premiums.

but eventually actuarial certification of the contribution tables was required. In 1948 the collection of annu- itants' data began. In 1937 two sets of tables were produced.33 The societies were clubs. The 1992 set of tables produced by the Bureau was based on over 15 million lives exposed to risk—a far cry from the 84000 policies covered in 1843! Sickness tables and Friendly Societies By 1835 at least a million people were members of Friendly Societies. This led to a need to have actuarial tables for the purpose of valuing the benefits to be provided. In 1954 the Committee investigated the effect of individuals holding more than one policy. The provision of death and sickness benefits was common and was regarded as important.34 The societies usually restricted themselves to a par- ticular locality: they had a strong social element and typically had between fifty and a hundred members. and the resulting life tables included projection factors to allow for future improvements in mortality. In 1990. usually for working men.94 H I S T O R Y OF A C T U A R I A L TABLES the new method would be more satisfactory in practice. leading to the dis- solution of the society as the members grew older. assurance data would therefore cause the cost of annuities to be set too low. Modern mortality tables32 are now produced regularly by the Bureau under the guidance of a Committee appointed by the actuarial profession. Originally the weekly contributions were usually fixed rather arbitrarily and often proved inadequate for the purpose. . one for heavy mortality and one for light. In 1924 the continuous collection of data began and a permanent Continuous Mortal- ity Investigation Bureau was formed. The Standard Tables Program is updated regularly and is still used today. The original data for the continuous investigation were provided by 60 offices. though some women's clubs also existed. The mortality of individuals taking out life annuities is usually lighter than that of assured lives. From 1964 an inves- tigation according to cause of death was often included. and using tables constructed from. to the delight of many actuaries. Female lives were not included and each policy was assumed to be held by a different individual. but was usually subsidiary to the society's social function and the meetings were often held at public houses. Female lives were included in the collection of data from 1973 onwards. the Committee commissioned the production of computer software to generate most of the existing basic tables.

Multiple decrement tables and pension funds A life table shows a hypothetical population of births which is thereafter diminished by one decrement only. however. The largest was the Manchester Unity Friendly Society. In real life. All in all. showing a higher rate of sickness in the city. H I S T O R Y OF A C T U A R I A L T A B L E S 95 Although Richard Price35 had done a little work in this area in 1789. The report contained sickness tables divided by town. with appropriate adjustments.893—7. this was not a firm enough foundation for the purpose. bad news for friendly societies. Unfortunately the resulting tables were found to understate the actual rates of sickness experienced by many societies and the tables were withdrawn from use in 185I. . today. city. covering some 800 000 members.John Finlaison suggested that one reason for this problem was a mistreatment of withdrawals. In 1903 Alfred Watson (a distinguished actuary who later became Government Actuary and was knighted in 1915) published the results37 of a compre- hensive investigation into the Society's sickness experience during 1. Each of the Society's branches had to complete a set of cards. providing details of their members and the sickness they had experienced. In 1824 the Highland Society of Scotland collected data from 73 friendly societies and computed sickness rates. Over 3500 sets of cards were painstakingly collected.36 This was the first investigation into sickness rates using real life data. The Board of Directors had agreed to look into this and a new method of compiling the rates was developed. and rural areas. Watson began his 1903 report by explaining that the experience of the Society had become some- what higher than the rates determined from the last investigation. There were a few very large friendly societies with branches in many towns and villages. which were incor- rectly regarded as exposed to risk for a whole year in the year of withdrawal. Rates were also sub- divided by age and occupation and included the average length of claim and the probability of becoming sick again after recovery.1856-60. and 1866-79. There has been no subse- quent major investigation into sickness absence among members of friendly societies and the result is that the Manchester Unity tables are still used. The picture revealed by the investigation was that sickness rates were rising and mortality rates falling. namely death. Investigations of the same Society's experience had previously been carried out using data for 1846-8.

Bernoulli38 adjusted Halley's table for this purpose in 1760. Multiple decrement tables were ideal for the purpose of carrying out these valuations. The prospective liabilities had to be revalued every five years.40 Two actuaries called Manly and Thomas wrote the first comprehensive paper41 on the valuation of staff pension funds in 1901. i. which shows the number of survivors in the popu- lation at each age. and retirement table. so that a check could be made that there was enough money to cover future pension payments. multiple decrement tables really came into their own following the development of funded occupational pension schemes in the second half of the nineteenth century. in relation to a superannuation fund for civil servants. and how many would marry at each age. This started with 20000 active employees aged 15 and showed how many of them would die. They gave a multi- ple decrement table which they called a combined mortality. as well as death. Multiple decrement tables were first developed to investigate the effect that smallpox was having on mortality. Appropriate assumptions need to be made about the probabilities of death. The methods used to estimate how many people die at a certain age can also be used to estimate the number of members leaving a pension scheme at a certain age. withdraw (for example.96 H I S T O R Y OF A C T U A R I A L TABLES populations are often diminished by other causes as well. through . and retirement at each age: these can be checked against the scheme's actual experience after it has been in operation for a few years. In 1862 Wittstein published a paper59 mentioning independent and dependent decrement rates. resignation. Such situations can be represented by a multiple decrement table. These monetary functions fulfilled the same purpose as their counterparts in life tables. For example. However. The first recorded use of a scale relating salary to age was in 1821. A popula- tion of employees might be reduced by resignations. a popu- lation of spinsters would be reduced by marriage as well as death. Using given probabilities of marriage and death at each age. after allowing for the various decrements at earlier ages. ill-health retirements. to ease the burden of calculation. An additional feature was that the monetary functions for pen- sion schemes for salaried staff often incorporated an increasing scale to take account of expected salary increases in future on promotion. A further development was the preparation of monetary functions based on the multiple decrement table. and age retirements. subdivided between married and unmarried. he derived a table starting with 10000 unmarried women aged 16 and showed how many of them would survive at each age in the future.e. withdrawal.

The table was based on assumed probabilities. The history of actuarial tables shows that their makers have. Particular atten- tion has been paid to accuracy. which is not so considerable as to need any apology. though. since even quite a small error could cost someone quite a lot of money. no doubt puzzling the schoolboys it was meant to educate. who are 'men utterly broken down in the service. by and large.' On the whole. would survive in service at each future age.What made it worse for him was that the errors had been pointed out by Weyman Lee. stated on the title page:6 'Examined also and corrected at the press by the Author himself. Errors could be caused either by the author or the printer. users of actuarial tables have had good reason to be satisfied. who are generally suffering from brain trouble or some chronic incurable disease*.' An edition of Robert Recorde's The ground of arts in 1632 contained a very erroneous table of the value of an annuity certain. which he had published in 1730.42 John Richards43 had to admit in 1739 that there were errors in the tables of annuity values on three lives contained in his book The gentleman's steward. and hence how many of them. many improvements were introduced over the years to make things easier for users—and in Victorian times it even became usual to give the logarithms of some of the functions tabulated. but hope that few of them amount to so much as an unit in the first place of decimals. Monetary functions were derived. in order to . a writer whose work was generally of poor quality! Richards corrected the worst errors but could hardly have inspired confidence in his readers when he went on to say: 'I am aware that a great many small mistakes of this kind are crept into the Tables. looking for accurate. in a form which they can easily use to carry out the calculations required. H I S T O R Y OF A C T U A R I A L T A B L E S 97 resignation) or retire each year in future. Meeting users' needs Users of actuarial tables are. and that affects it to but 1/10 of a year's value. tried to respond to these needs. The authors allowed for mortality being higher among early retirements. in an endeavour to reassure purchasers of his book on leases and annuities. Richard Witt in 1613. relevant and up-to-date figures. which ran many of the pension funds then existing in Britain). based on a pension age of 65 and an interest rate of 4% per year (the rate paid or guaranteed by the railway companies. based to some extent on the experience of existing pension funds. As we have seen. of course.

98 HISTORY OF A C T U A R I A L TABLES avoid having to look them up in logarithm tables. 3. tables no longer needed to incorporate logarith- mic equivalents. Following the advent of desktop electronic calculators and the introduction of suitable calculation software in the second half of the twentieth century. the number of centenarians in the population is increasing rapidly Fig. However. Nevertheless they still perform a vital role in the finances of life assurance companies and pension funds. Finally. High Holborn. and new tables are still needed every few years to cope with the latest mortality changes. Brilliant minds devoted themselves to the search for approximations. it became unnecessary to tabulate monetary functions.) . short cuts. Actuarial tables are therefore nowadays much simpler than they were a hundred years ago. with the spread of mechanical desk-calculators in the first half of the twentieth century. (By Kind permission of the Institute of Actuaries. London. and laws of mortal- ity.5 The Institute of Actuaries.

edited by Steven Haberman and Trevor Sibbett. banking. and pub- lished by Pickering & Chatto in 1995 in 10 volumes. D. volume VII. B. alternative risk transfer. risk management. 1873. J. 3.24). columns 976—1016. Wiley. Dover. published by Wiley in 1990. Leyden. Struik. Pickering & Chatto. 1936 (publica- tions of the Mediaeval Academy of America. in fact. S.J. arts and sciences division. Some additional books of tables are described in the article entitled 'Table' (written by Augustus de Morgan?) in Tlie English cyclopaedia. 1571 and later editions. 1965 (originally published 1954). one of the present authors (Christopher Lewin) has written Pensions and insurance before 1800: a social history. New York. including the origin and graduation of life tables.. L'arithmetique. by Anders Hald. Cambridge. 1585 (also in a col- lected edition of 1634). On a somewhat lighter note. 1. H I S T O R Y OF A C T U A R I A L T A B L E S 99 and Finlaison's question of whether there is an inextensible upper limit to human age has yet to be answered. Given the long history of actuarial science. edited by Steven Haberman and Trevor Sibbett (10 vols. and credit derivatives'—any field. no. Trenchant. F. it is perhaps appropriate that the Institute of Actuaries is located behind one of the few surviving Tudor frontages in London. 1990. . Pegolotti. gen- eral insurance. such as capital project appraisal. 2. 5. Notes In the notes below 'H & S' refers to History of Actuarial Science. edited by Allan Evans. 1566. It is gradu- ally expanding into new areas of work. which requires knowledge of both risk and finance. Lyons. A concise history oj mathematics. La pratica della mercatura. Hald. New York. London. The useful preface by Professor Haberman gives more detail on some of the topics covered here. A. A history of probability and statistics and their applications before 1750. 1995). La pratique d'arithinetique. London. Mass. Stevin (of Bruges). 4. c. Another very worthwhile technical book is A history of probability & statistics and their applications before 1750.This contains reprints of many historic actuarial tables and accompanying text. the profession is not standing still. However. Further reading Perhaps the best book for further technical reading would be the History of actuarial science. healthcare systems. 1558. which will be published by Tuck well.

in order to estimate annuities &c.3. Philosophical transactions 17(1693). and existing in manuscript in the library of the Institute of Actuaries.165-84. with their grounds expressed infoure tables of frac- tions. table and 654-6. Kersseboom. J. with an attempt to ascertain the price of annuities upon lives'.. touching the buying or exchange of annuities. The doctrine of annuities and reversions. A letter from Mr. the third column started with 170 and each subsequent number was probably obtained by multiplying die number immediately above it in the column by 189 and adding the number from the first column in the current row. which can also be expressed as 19/170. H & S. Inleiding tot de Algemeene geographic. is in H & S. for example. Edmund Halley und Caspar Neumann. J. drawn from curious tables of the births and funerals at the city of Breslaw.. 197-205. R. N.2. The Hague. 1. 1883.. I. F. 1725. The main compound interest func- tions were then obtained by dividing a number in one of the columns by the number in another column in the same row.. February 1738. 596-610.) The working method used to calculate this table was to tab- ulate three columns: the first was 170 raised to the successive powers 1. An English translation of the Appendix. Tables ofleasses and interest. Simpson. 10. Halley. Graunt. W. London. re inclosing tables extracted from ye bills of mortality for the last ten years. Smart. H & S. De Moivre. This is an English translation. of works originally published in Dutch. J. London.. 23-127. H & S. London.100 H I S T O R Y OF A C T U A R I A L TABLES 6. 7. 1-125. Annuities on lives. H & S. 8. 12.'An estimate of the degrees of the mortality of mankind. Smart to George Heathcote Esq. 1628. 15. the second was 189 raised to the same powers. Amsterdam. I. 1748. E. Graetzer. 13. T. 1742. Arithmetical! questions. (The reason for using such a peculiar rate of interest was that it corresponded approximately to the rate which would have been obtained if a 21-year lease was bought for a capital sum equal to 8 times the annual rent. The First Treatise is designated the Third Edition and is dated 1748. 1. and from them shewing the probabilities of life. 1662. Witt. Breslau.. Rickman in 1886. . Halley rather misleadingly mentions 1238 births in a supplementary paper but this figure should not be taken as the number which is consistent with the numbers in his life table. was based on an interest rate of 11. By Mr Aecroict? 7a. Struyck.. One of the tables. A.Translation into English in Institute of Actuaries Library. Anon. 1613. III. the other two treatises are dated 1742. 11. London.4.. 14.18%. entitled Appendix to the conjectures on the state of the human race and the calculation of life annuities. made by E. Essays in political arithmetic contained in three treatises. London. 1740. Natural and political observations made upon the bills of mortality. 9. 207—41.

'A sketch of an analysis and notation applicable to die estimation of the value of life contingencies'. London.'. with a commentary. Weidemanns Erben und Reich. 119-91.V 79—143). Milne. Also Gompertz. London. 1. and on a new mode of determining the value of life contingencies. I. An English translation is in die Library of die Institute of Actuaries. "On the nature of the function expressive of die law of human mortality.8. An English trans- lation appears in H & S. Wargentiii. 11. H & S. London. 357—86. 1782. H I S T O R Y OF A C T U A R I A L T A B L E S 101 16. IX. J. 22. 4th edn. R. Ogboru also explains how he thinks a modified version of Dodson's life table was calculated. Philosophical 'Transactions. House of Commons. 230. Benjamin. Essai sur ks probability de la durfe de la vie humaine. A treatise on the valuation of annuities and assurances. 2 vols. Life tables founded upon the discovery oj a numerical law regulating the existence of every human being: illustrated by a new theory of the causes producing health and longevity. H & S. 216-22. 30-43. vol. 1829. B. 12-69. 24. 3. Edmonds. Extracts from the manuscript were printed in an article by M. H & S. Philosophical transactions 116 (1826). ER. 243-9. Young. H & S. H & S. 23.79-118. IV. N. 1825. Calculations deduced from first principles. Leipzig. A typescript of a translation of die manuscript made in 1984. J. 'The actuary in the eighteenth century' in Proceedings of the centenary assembly of the Institute oj Actuaries. on the evidence and elementary facts on which the tables of life annuities are founded.for the use of the societies insti- tuted for the benefit of old age. H& S. 1832. A. 136. London. 18a. Gompertz. 19. M. Finlaison. 1746. Mortality in Sweden according to the 'Tabell-Verket' (General Register Office). 1-24. 216-85. 1785. T. Translated into English for the ninth international congress of actuaries by the Thule Insurance Company in 1930. J. Life annuities—Report of John Finlaison. actuary of the national debt. H & S. by a member of one of the societies. Dale. Deparcieux. is held in the Library of the Institute of Actuaries. 1-17. London. 513-83.Tetens. J. R.P. W. E. pp. Dodson. Paris. 1772. 1766. London. Vol. Ogborn.S. 286-92. Philosophical transactions 110 (1820).. 17. H & S. Leipzig. IV. . II. 21. . 1950.. First lecture on insurances. I. Observations on reversionary payments. 88-103. 1815. T. 1756 (H & S. 192-215. H & S . 20. 25. pp. pp. 'A formula for expressing the decrement of human life. Bart. In a letter to Edward Hyde East. II.. Price. II. In a Letter to Francis Baily*. P. II. Introduction to the calculation of life annuities for life and reversions which depend upon the life and death of one or more persons. II.

409—30 deal widi the contributions required to provide sickness benefits. Tables exhibiting the law of mortality. Edinburgh. 'The mortality in companies with successive entering and exiting members'. 'On the law of mortality and the construction of annuity tables'. Investigation of mortality in the Mian army. An example is 'Mrs. W. pp. 36. D. 25-178. near Dudley. 91-113. X. 'An attempt at a new analysis of the mortality caused by small- pox'. B. Bernoulli. and the adjustment of numerical tables (Part III)'. 29.Jones. Worcestershire—a lodge book from the 1860s survives in the collection of one of the authors. 'On interpolation. Journal of the Institute of Actuaries 12 (1866). 37. 39th volume. Nottingham. 301-10. 1971. 166-208. Woolhouse. X. H & S. Manchester. 30-78. 30. 31. 1862. Journal of the Institute of Actuaries 8 (1. London. Griefswald. Report on friendly or benefit societies. which appear to have taken place at Cooksey's hotel. Elderton. Histoire de I'Academic Royale des Sciences. summation. 27. as deduced from returns by friendly societies in different parts of Scotland. exhibiting the law of sickness. X. 34. W. London.VIII.A. 5th edn. 32. Makeham. A treatise on friendly societies. 324-34. . Richard C. Koch. X. London. English translation by Trevor Sibbett in H & S. S. T. W S. R. London. numbers 1—19. London. 219-300. Institute of Actuaries and Faculty of Actuaries. 1-24. 1973-2000. H & S. Archiv for Mathematics and Physics with special attention to the needs of Teachers in higher educational Institutions. C. London.859). A. deduced from the combined experience of seven- teen life assurance offices. 35. H & S. 1835. 1903. 23. J.102 H I S T O R Y OF A C T U A R I A L T A B L E S 26. Watson. The Chib paid benefits on death and maternity and also hired a surgeon to attend members in their confinements. B. 33. 38. run by Mrs Ruth Cooksey. 39. 1766. 1893 to 1897. W. A great deal of ale was consumed at the meetings. 132-46. 1839. Highland Society of Scotland. Ansell. Bradley. A translation into English from L. H & S. which was located at Rowley Regis. X. 1824. Cooksey's Womens Club*. 1843. read 1760. 1843. 260-72. H & S. 'Notes on the construction of mortal- ity tables'. 1—37. C. H & S. A series of tables of annuities and assurances calculated from a new rate of mor- tality amongst assured lives. is in H & S. H & S. Price. Sickness and mortality experience of the Independent Order of Oddfellows Manchester Unity Friendly Society. IX. Continuous mortality investigation reports. IX. II. 136-76. and Fippard. Woolhouse. Paris. Wittstein. H & S. Observations on reversionary payments. W. Smallpox inoculation: an eighteenth century mathematical controversy. IX. 1792.Journal of the Institute of Actuaries 56 (1912). 19-29. P.

169-249. and Annuities on lives. 'On the valuation of staff pension funds: with tables and examples'. H & S. 101-68. teaching the perfect uvrke and practice of arithmeticke. H & S. H. H I S T O R Y OF A C T U A R I A L T A B L E S 103 40. The gentleman's steward and tenants of manors instructed. 1730. London. 43. London.VI. Observations on a superannuation Fund proposed to be established in the several pub- lic departments by the Treasury minute of the 10th August 1821. Recorde. Manly. and Thomas. . The ground of arts. Ernest C. 1739. 42. Richards. 209-87. R. London 1632. J.W. revised by Robert Hartwell. of Exon.VI. 41. Journal of the Institute qfActuaries 36 (1902).

Log Cos. .Log Sin.

4 The computation factory: de Prony's project for making tables in the 1790s IVOR GRATTAN-GUINNESS A large set of logarithmic and trigonometric tables was produced at the end of the eighteenth century under the direction of Gaspard Riche de Prony.) . despite the fact that printing was started more than once and various efforts were made over the years to find finance.1 De Prony's ambitious tables proved too voluminous and expensive for commercial publication. They eventually saw the light of day in a much reduced form over fifty years after his death.1 Although they were completed in \ 801. 4. (Courtesy of the Tomash collection. The title page does not even mention his name. their size made publication a costly task and it was never done. Fig.

I shall refer to him as 'de Prony*. From 1798 he was also made Director of the Ecole des Ponts et Chaussees. at the newly founded Ecole Poiytechnique from 1794. (Courtesy of Ivor Grattart-Guinness. another post which he retained until the end of his life. 4. and he added this name to his own. de Prony joined the associated engineering corps but maintained connections with the school. After studying at the Ecole des Fonts et Chaussees in Paris in the late 1770s. In the reforms of educational institutions fol- lowing the Revolution of 1789 he was appointed professor of analysis and mechanics with Joseph-Louis Lagrange (1736—1813). de Prony's career: education and mathematics for the engineer Gaspard Clair Francois Marie Riche de Prony (1755—1839) was born in the Beaujolais region of Southern France to the Riche family. . They owned land at a location nearby called Prony.2 De Prany was sketched in one of his student's notebooks at the Ecole Poiytechnique in 1802 or shortly afterwards. This project also turned out to be influential to Charles Babbage's work. transferring to a post of graduation exam- iner in 1816.) Eventually a reduced edition of some tables appeared in 1891. in conformity with his own preferred form in his many publications. but his younger brother Claude used only the original surname during his career as a distinguished naturalist.106 THE C O M P U T A T I O N FACTORY Rg.

He wrote a few treatises. which was partly born out of the reform of weights and meas- ures then being undertaken. citizen de Prony was appointed chief engineer to the departement of Oriental Pyrenees. therefore. a large number of papers. and for a simple dynamometer (the 'de Prony brake') which he introduced in the late 1810s. de Prony is remembered today only for his tables. some short pamphlets. when he was in his 36th year. THE C O M P U T A T I O N FACTORY 107 He was elected in 1795 to the new classe des sciences mathematiques et physiques of the Institut de France. Perhaps opportunely. to prepare a detailed map of France to facilitate the accurate measurement of property as a basis of taxation. but he announced his desire to stay in Paris.2 He made a stark contrast to his colleague professor Lagrange. where a scientist or administrator held several posts at once. such as the directorship of a small Ecole de Geographes which he ran from 1795 until its closure in 1803. In 1801 he was appointed to a supplementary position at the Bureau des Longitudes. For example. Such was the style of professional life in the Paris of his time: the so-called 'cumuT system. . and also some subsidiary ones at various times. early on in the new decade de Prony set up a Bureau de Cadastre in Paris. and he retained this position in 1816 when that institution was restored after the fall of Napoleon.3 In connection with this plan. and in many related areas of science. He brought to much of his work a philosophy which bore strongly upon his tables project: namely the desire to render theories math- ematically in ways which were both amenable to numerical calculation and accessible to observational data. publishing it at length in the school's Journal. the learned body which replaced the old Academic des Sciences. He held all these posts throughout his life. who preferred a theoret- ical algebraic approach to the calculus of his own devising (although the 'Lagrange interpolation formula* belongs to this time also). The tables project In 1790. for his numerous activities in engineering practice and education. but in his lifetime he was very well known. This predicament may have inspired him to launch the tables project. at the Ecole Polytechnique he included an optional but substantial course on difference equations and related topics. and published several editions of his lectures on mechanics at the Ecole Polytechnique. it was decided that a very large set of logarithmic and trigonometric tables would be prepared.

connected information is hard to find.108 THE C O M P U T A T I O N FACTORY From the mid-1790s the Bureau seems to have been housed with de Prony's Ecole de Geograpb. including the mathematicians A. [. nor have I traced its archive (if it still exists). I conceived all of a sudden the idea of applying the same method to the immense work with which I had been burdened. No useful information exists on the organisation of the work-room (which must have been one of the largest in Paris at that time) or the full personnel or the budgets for the project.4 I have not been able to find even the address of the Bureau de Cadastre. I have reason to believe that I had already been prepared for this conception by certain parts of mathematical analysis. and apparently the waste sheets used for producing or checking the tables were discarded after use. C. As he later recalled. including A. M. M. and so on. as well as it can be traced. Parseval (of the well-known formula in infinite series) and the text- book \vriter J.6 The personnel were divided into three sections according to the work they did. The mathematical details of the tables project will be surveyed in the next section: the chronology is described here. and the differences . Given these provisos. and also considered the choice of initial values of the numbers or angles. as an example of the great advantages of this method. the number of decimal places to be adopted in each table.] I came across the chapter where the author treats of the division of work [fratw'fj. who determined the values. de Prony devised the project explicitly following the principles of the division of labour which had been laid down by Adam Smith in A treatise on the wealth of nations of 1776. The second section comprised several 'Calculators'.5 de Prony gave some details of the project in a 'Notice' read to the classe des sciences mathematiques et physiques in 1801. Legendre. A. This qualifier needs some explanation.es. the former two were also involved with the reform of weights and meas- ures. and had responsibility for the calculations of the navigational tables required by the (newly formed) Bureau des Longitudes for publication in their annual journal Connaissance des temps. and latter two also acted as influential political figures. Gamier. They chose the mathematical formulae to be used for calculation and checking. here is brief outline of the history of the tables. the manu- facture of pins. and some sources contradict details given in others. The first section contained a handful of mathematicians. G. and Lazare Carnot. Despite the fame (or notoriety) of the project. on which I was then lecturing at the Ecole Polytechnique. citing... soon after it was finished. Prieur de la Cote d'Or. and to manufacture loga- rithms as one manufactures pins.

one of the most hated symbols of the ancien regime had been the hair-styles of the aristo- cracy. so that each set could be checked against the other. some of the equipment that had been used (presumably on the Fig. 4. in that two sets of each table were produced from different equations. They also prepared a page of tables for the numerical work by laying out the columns of the chosen values and the first row of entries. The project was run twice. THE COMPUTATION FACTORY 109 of various orders.Thus these artists were converted into elementary arithmeticians. it was returned to the second section.8 When a page was completed. and preparing the instructions on the preparation of the remaining entries on the page.3 Hairdressers to the aristocracy were among the economic victims of the French Revolution. By 1794 seven hundred results were being produced each day. De Prony trained them in elementary arithmetic and set them to work as production line labourers in his table-making factory. executing only additions and subtractions. to check the figures using formulae chosen by the first section. .10 In addition.7 Many of these workers were unemployed hairdressers. to its most simplest expression' left the hairdressing trade in a severe state of recession.9 After the project was completed in 1801 several of the calculators were transferred to the Bureau des Longitudes to carry on similar work there. and the obligatory reduction of coiffure 'as the geometers say. that needed to be calculated. a large team of between 60 and 80 assistants. These calculations were done by the third section.

1832. Green. London.000. his policy of using large numbers of decimal places. on the application of inadtintry to the purpose of (tilcultiting and printing mathematical tables. It seems that Babbage had already envisioned his idea of mechan- ical calculation before learning of de Prony's tables.2 Clearly he was struck by de Prony's production of the tables following an industrial process. 4. London. . by acquaintance with that project. 1835. 241-5O. file BAB U21. recognise difficulties. and was hoping to imitate the process by mechanical means. C. Farnborough. Indeed. but he must surely have heard of die project. TMa of logarithms of the natural numbers from 1 to 108. London. working with sequences of differences of various orders is reminiscent of de Prony (although of course it was not novel with either man).. the economy of nutmtfictUKs and niadiitiery.1 The latter date coincides -with Blagden's involvement in the possible printing of the tables: Babbage does not refer to de Prony here. he rehearsed some of the same material on de Prony's project and on mechanical calculation. Babbage. Babbage. C. but he may well have been helped to crystallize some of his ideas. Babbage. and lie may have realised that mech- anization was the better way to produce tables of this size. in a pamphlet of 1822 on 'the application of machinery to the purpose of calculating and printing mathematical tables'—his main publication for securing governmental support for production of the Difference Engine—he gave some account of de Prony's project and noted that mechanical methods would speed up the process of calculation. 1864. 1969. Booth. Mss. 37182. and so on. London.Add. arts. Of. soon afterwards. p. Notes . vi. Mawman. 1st edn. pp. but he examined the Observatoire set of tables during his visit to Paris perhaps in 1826 and had copies made of some of de Prony's writings on the project. Moreover. Longman. C.37183 and 37201. Knight. one from 1812 and the other from 1819. 1. A letter. Babbage does not seem to have met de Prony or corresponded with him. 1827. 4th edn.. 3. reprinted Gregg. 42-3.4 Thus his contribution was to substitute manual labour by engineered automation in the construction of tables. 2. Passages from the life of<t philosoplter.110 THE C O M P U T A T I O N FACTORY de Prony's influence upon Babbage Charles Babbage (1791—1871) gives in his autobiography two occasions for the origins of his interest in mechanical calculation. In his book on manufactures. C. and in his correspondence held at the British Library. Pertinent materials arc in the Babbage archive at the Science Museum (London). Babbage.3 The nature and extent of de Prony's influence on Babbage needs some exploration. 1822.

and estimated that a volume of 1200 large pages would result. Nevertheless. but Blagden's death in 1820. But financial crises in France began to intervene. and following the fall of Napoleon. The publishers Firmin Didot were charged with the task of printing. having served previously in the rotating post of secretaire. scuppered these hopes. who had died in 1799" and as part of his responsibility at the Bureau des Longitudes he was involved in producing solar and lunar tables for navigators. Borda. collaboration was sought with the British government for financial support. He was also a leader of the expedition in the mid-1790s to triangulate the longitude between Paris and Barcelona. THE C O M P U T A T I O N FACTORY 111 preparation of the map) went to the Depot Generale de la Guerre. At that time de Prony read his 'Notice' to the dasse. at a cost of 80000 francs. especially through the good offices of the physician Sir Charles Blagden.13 At the end of that decade. hopes remained alive for many years to com- plete the work. and printing stopped.12 apparently about 500 pages were in proof by 1802. a leading practical astronomer of his time. together with the introduction. for in 1803 he was appointed founder secretaire perpetuel of the classe for the (so-called) 'mathematical' sciences. and the British desire to have the tables converted from the original centes- imal system of angular division to the sexagesimal system. Delambre was to be an influential aid to de Prony. Firmin Didot announced the volume (s) in some of their cat- alogues of the 1810s. a milit- ary organization concerned with geodesy and cartography. His pieces and de Prony's 'Notice' were reprinted in the Memoires of the classe in 1804. C. a massive task in itself. He gave the project some publicity in one of his regular reports of the activities of the classe. Delambre himself was much concerned with the compilation of tables: by 1801 he had com- pleted a much smaller set of logarithmic and trigonometric tables initiated by the mathematician and instrument designer J. One of de Prony's supporters was Jean Baptiste Joseph Delambre (1749—1822). who was then a repres- entative of the British government in France. He revived interest to the extent that printing seems to . In pamphlets and notices of the 1820s de Prony announced the possible collaboration to the Academic des Sciences (as it had reverted to) and he hoped that the 'unique monument' could be kept in its original form. and the task was set to find money to finance the printing. writing up the details in three large volumes. Legendre published some entries from the logarithmic tables in his treatises on transcendental and elliptic integrals. whereupon de Prony added an historical note on the errors found in previous tables.

and Firmin Didot formally returned their set to the Bureau des Longitudes in 1833. de Prony himself retained the other set. and that set was placed in the library of the Bibliotheque de 1'Institut. by Pierre Alexandra Francisque Lefort(-Latour) (1809—1878). Many treatises appeared from commercial houses such as Firmin Didot. Didot must have lost much money as a result of the failure to publish. Biot reported his finding to the Academic des Sciences. they also supported a large number of journals. Seemingly they were able to take the loss. sup- ported research both by offering employment and establishing journals.16 Towards publication Both sets of the manuscript tables are still preserved. there then follow (not necessarily in this order) a volume of sines of parts of the radius. Several other national organizations. and this capability reflects the astonishing scale of publishing in science and technology in France at that time. B. in the posses- sion of the niece's daughter and son-in-law. but his niece and heir did not include it in his manuscripts that she sent to the Ecole des Fonts et Chaussees. The first eight volumes contain logarithms of numbers from 1 to 200 000 calculated to 14 decimal places. One is held in the library of the Paris Observatoire. especially military ones. in the Observatoire. Lefort was a graduate of the Ecole Polytechnique and the Ecole des Ponts et Chaussees (during the time that de Prony had been respectively graduation examiner and Director).112 THE C O M P U T A T I O N FACTORY have begun again. often of 251 folios each (containing the tables for a run of 25 100 values of the numbers or angles). in the course of making an extensive study of the other set. The others are large-format tomes. hopes still remained for printing. three volumes each of sines and of tangents (including their . the last one of which consists of an account of the methods of calculation (and a few basic tables). It was found only in 1858. In the new phase of school and higher education the Ecole Polytechnique and the Ecole des Pone et Chaussees were only two (import- ant) institutions among a national network which no other country then matched.15 but they were not realized.14 After de Prony's death in 1839. Only a few printed pages seem to have survived. Each set comprises 19 volumes. Firmin. and then a career engineer in the Corps des Ponts et Chaussees. but once more it ceased. Thus someone in applied mathematics and mechanics might have up to 40 outlets for his papers by the 1820s. His grand- father-in-law J.

4 De Prony produced two independent manuscripts of his tables in order to check one against the other. Instead of building up the final values from differences of various orders and by interpolative formulae. and gives a less complete account. the last volume in the Institut was originally with its brother in de Prony's family: it was transferred from the Observatoire some time after 1858. In fact. and the product of sines with cosines. only the Institut set has this last volume: presumably the other set had had a copy also. THE COMPUTATION FACTORY 113 Rg. In the 1980s John Fauvei showed them in an Open University video for a course on the history of mathematics. In the mid-1870s the Scottish mathematician Edward Sang (1805-1890). and the differences pertaining to succeeding values (x + It) were calculated and inserted into an adopted formula to produce the required values of/(x+ It). hut it is now lost. (Courtesy of the Open University. further. ini- tiated a debate with Lefort about some of de Prony's procedures as Lefort had recently described them. One of these nineteen folio volume sets is now In the Bibliotheque de I'lnstltut in Paris. who had recently produced some extensive logarithmic tables of Ms own. the exact values off(x) were calculated for some initial values x from a known formula (usually a series expansion). A main motive for this strategy was to avoid multiplica- tion and division and to reduce the calculations to sums and (especially) .17 As de Prony had stated. 4.) logarithms for some ranges of the argument). his design of the tables was like an inverse to normal methods. one volume each of the ratios of arcs to sines and of arcs to tangents.

His daughters fought a long and hard but ultimately successful battle to deposit the manuscript with the Royal Society of Edinburgh.114 THE COMPUTATION FACTORY Rg.5 In contrast to de Prany's computation factory the Scottish mathematician Edward Sang's table-making project was a cottage industry with his daughters as the only other workers. 4. De Prony used two different sets of formulae for the two sets of tables for checking purposes. In the design some obvious moves were missed. Lefort showed that the pos- sible errors obtainable were not as large as Sang feared. (Maybe he had used it but regarded it as too obvious to need stating. differences.) He also criticised the strategy of rounding or truncating the differences of the highest orders (which he called a method of'vitiated differences'). as Sang pointed out. such as deploring the apparent discard of the loose . But like de Prony Sang died with 47 volumes of his 28 and 15 place tables unpublished. de Prony did not employ the obvious relation either as a means of reducing or simplifying interpolations or as a check on the decimal parts of the logarithmic tables. His repertoire of techniques was also determined by the perceived need to calculate the entries to so many decimal places (the numbers varied between 14 and 29). on the ground that rounding or truncating errors were prematurely introduced already in the values found at the higher orders of difference and therefore cumulation could occur upon transmission back to the lower orders. For example. which the hairdressers could fairly be expected to handle. but he had his own criticisms of the project.

played an important role: {in terms of the definition of A«(. These computations proceeded up to the column of differences (prepared by the second section) which contained (almost) equal figures in each row. when of course the process stopped. the logarithms of prime numbers were calculated from an expansion which can be written as: w Further terms were deliverable via expansions such as Logarithms of compound numbers were produced from those of their prime factors by addition. where the forward difference was defined as usual: An iterative formula on orders of differences on neighbouring values of func- tion.v) in (2)). Among special formulae used. avoiding multiplication and division.1° and b in steps of 0.01°. where it took values in steps of 0. (The quadrant was . Mouton. and addition formulae for sines and cosines were deployed for intermediate values. The names of the required differ- ences were laid out in columns on the page by the second section for the hairdressers to calculate as per instructions in the body of the page. due to the seventeenth-century astronomer G. de Prony devised formulae expressing expansions in terms of differences of various orders. for example sin(<i + b). For the trigonometric tables the standard power-series expansions were used for the initial entries over 10° intervals. THE C O M P U T A T I O N F A C T O R Y 115 On de Pretty's mathematical methods So that the hairdressers could be called upon simply to add and subtract. One of these was a theorem due to Euler.

Some check formulae were used. Up to 50° difference formulae were deployed. these were expressed as polynomials in cot x. For differences of various orders. results such as were used. A favoured one for the sine formulae was Eider's: . For convenience the usual symbol for degrees indicates these angles here. Log tan tables were produced from differences such as and various similar formulae derivable from it. with (ax/mi ax) expanded in power series of ax.116 THE C O M P U T A T I O N FACTORY On de Prony's mathematical methods cant divided centesimally. which was small when x was dose to 100°. For log sine tables. and for intermediate values The tangent tables were devised by various formulae. then up to 94° recourse was made to and for the last few a power series in cot x (which would then be taking very small values) was used. some use was made of Lagrange's symbolic generalized version of Taylor's theorem: This formula required knowledge of the derivatives ofj(x): for log sin x. rather similar in form to Mouton's (3): for extreme values of x in the range.) For higher-order differences de Prouy adopted 'the extremely elegant and simple formulae proposed by the citizen Legendre*.

even if fairly frequent in places (and written in red ink above the black originals). Lefort reports detecting 'arbitrary corrections* in places in the tables. and expressed himself with careful vagueness to Blagden in 1819 when the col- laboration with the British Government was in the offing. The collation of the two sets was completed in 1862 by officers of the Depot General de la Guerre (the last volume of the set is so annotated and signed). cos.21 The year of publication. the manner of compilation seems to have been corrigible. and may not have been noticed. In contrast to his own completion of the Borda tables. One aspect of the tables which is not too clearly described is the effect of errors of calculation. of the (truncated) logarithmic.!S What remains unexplained is the reason that de Prony chose to calculate these tables to such extraordinary numbers of decimal places in the first place. Lefort's reservations and Sang's criticisms carry weight. as a replacement to the Borda/Delambre tables. even though the Service Geographique de FArmee found few errors in the tables when preparing their version for publication. In his discussion of the tables. and disliking de Prony's method of producing the final values for the log sine table. and in the short . In a passage which excites suspicion. Lefort was doubtful of the value of publication. but they seem to be occasional. with the latter still rendered in centesimal angular division.2" Either by accident or influence. One would presume that the errors would (or could) iterate onto entries in the vicinity of their occurrence. By and large. tan and cot tables. THE C O M P U T A T I O N FACTORY 117 sheets upon which calculations had been made before transcription of the results onto the pages. 1891.This edition comprised 640 (printed) large quarto pages. or did de Prony want to keep the threat of the Pyrenees appointment of 1790 well at bay? Delambre knew the project well. which went only to six or seven places. but only in extraordinary cases. he opined that 'These tables will not serve in the usual cases.'" When de Prony's set had been reported to the Academic des Sciences in 1858. and log sin. Did the taxation system really require the cadastral survey to be such an exact science. but thought that a rounded-off version to eight places was desirable. but this detail was not mentioned in the preface written by the Director of the Service Geographique de 1'Armee. In fact. this volume did not even mention de Prony in its title (so that its exist- ence has understandably been overlooked by historians). thirty years later the Service Geographique de l'Armee of the Depot published just such an edition. hopes had again been expressed that printing could be achieved. was close to the centenary of the project.

He did not refer to the fine set of tables prepared for hydrographers by Valentin Bagay (1772—1851) and published by Firmin Didot. 5. Basel. 3 vols. and 1. 'Commemoration en souvenir de Prony (1755—1839)'. Notes 1. 3. see J. lettres). M. For the work of Borda and others on tables and on related projects of this time. It seems that few errors were found. Critical enquiry 21 (1994). 1919 (as Annales de I'IMversite de Ljon. Cartography in France. Edwin Mellen Press. 1660-1848. and mentioned the need for tables which go beyond the seven places given there. On the historical context. at least for the early places of the long computations. So congratulations are due to de Prony for his design of the project—and hats off to the hairdressers. Further reading The most substantial survey of de Prony's career is M. and 16. educator and scientist. R. 2 (droit. pp. His manuscripts and library—both massive collections—are held in the JEcole des Fonts et Chaussees in Paris. Queenston and Lampeter. Bradley. especially pp. Berlin (DDR). Lewiston.. La vie et les travaux du Chevalier Jean-Charles Borda (1733-1799). and Deutscher Verlag der Wissenschaften. Konvitz. 183—202. Grattan-Guinness.118 THE C O M P U T A T I O N FACTORY preface the Director reported that the Borda/Delambre tables were practic- ally sold out. 47—62. An earlier version of this . Pebereau. See also X. Lyon. 33). Chicago University Press. Lefebvre. sec. The author's archival researches carried out in Paris over the years were aided by grants from the Royal Society and the British Academy. 1998. 8. Bulletin de la Societe d*Encouragement pour I'Industrie Nationale 139 (1940). Convolutions in French mathe- matics. Chicago. A career biog- raphy of Gaspard-Clair-Fmngois-Marie Riche de Prony: bridge-builder. 1990. see J. fasc. is a moderately useful history of French cadastral sur- veys. 1987. Daston.s. 6. 580—93. Walckenaer (editor). 1953. 18'00-18'40:from the calculus and mechanics to mathematical analysis and mathematical physics. Birkhauser. On the context of the project in French cartography of the time. Mascart. Paris. especially Chapters 2. Le cadastre Jrattfais. where several of the same functions had been calculated to seven (or a few more) decimal places. 68—98.22 The rounded-off tables were transcribed from the Observatoire set and the proofs checked twice against the original. Herbin and A.'Enlightenment calculations'. too. see L.. n.

Konvitz. 1824.-J. des sciences mathetnatiques et physiques de I'lnstitut de France. 1824. and also a manuscript career doc- ument held at the Ecole des Fonts et Chaussees. Paris. There are also files in the Archives Natiouales. 3. 4. 2 (1796). as are the minute-books of die Bureau des Longitudes. where the supposedly spherical Earth was projected onto a tangent plane. 1832. F. F. 49. Baudouin. The quotation here. Baudouin. p. G. compare his friend the educationist and novelist Maria Edgeworth on him in Maria Edgeworth in France and Switzerland. pp. 1803—4. 209-73. Lefrancois de Lalande. For the map de Prony used a 'central' projection.107-71. cah.225—52 (also published in his Discours et lemons sur I'indmtrie. 1801. 1987. but according to his Instruction elementaire sur les moyens de calculer les intervalles musicaux. 193—216. Annals of"the History of Computing 1. C. Paris. Annales des sciences industrielles 16 (1824). M. Cartography in France. pp. 49-55). 149-209). F. pp. Also in his Methods directe et inverse des differences. Chicago University Press. Riche de Prony. 1979. Riche de Prony. but 1791 in Notice sur les gmndes tables. pp. Notice sur les grandes tables. Journal de I'Ecole Polytechnique. for a country of the shape and size of France. de Prony gives various dates for the date of commencement of the project:'an 2' (1793—4) in G. Didot. C. See footnote 7 for another example of inconsistency. 5. 9. C. These are the numbers indicated in G. 1796. (ed. Dupin. 53. comes from C. C. G. Riche de Prony. 'Introduction d'un nouveau cours [at the Conservatoire des Arts et Metiers^. 1801 (also in Memoires de la classe des sciences mathtma- tiques et physiques de I'lnstitut de France. Oxford. Baudouin. 6.2 (1. 2. 1786. Notice sur les grandes tables. See J. (1)1. Paris. Riche de Prony. Clarendon Press. between 150 and 200 assistants were involved. 4. C. Didot. p. For example. F. ms. vol. 7. Riche de Prony. cah. p. Notice sur les grandes tables. . Iraprimerie de la Republique. 177-85.990). 743-4. Didot. 151. Paris. M. 23. 1803. 2. Various other files in the Ecole des fonts et Chaussees are relevant (see J. THE C O M P U T A T I O N F A C T O R Y 119 chapter appeared as 'Work for the hairdressers: the production of de Prony's logarithmic and trigonometric tables'. cah. Paris. Chicago.803-4. 1-23. Notice sur les grandes tables. 'Suite des lecons d'analyse*. F. p. M. 4.172-282. pp. Paris. 459-569. and very proba- bly other files there (for example F14 2146 is quite useful): the very poor state of cataloguing has doubtless hidden other files. 3. Paris. 31-53. Paris. pp. 8. 1825. Paris. Imprimerie de la Repubiique. and the attached information. Bibliotheque astmnomique avec I'histoire d'as- tronomie depuis 1781 jusqu'a 1802. p. M. 4.49-55). G. 165—6). 1. M. F17 1393 and 13571. 1801 (also in Aietttoires de la classe. Bachelier. good representations of geodesies and areas were obtained. C. Colvin). 1660-1848. p. 67-8.

Imprimerie Republique. Lefort s obser- vations. A. 326—51. [138] reports having seen six incom- plete printed copies ot the sine tables. 14. 1824. Additions. Grattan-Guinness. Annales de I'Observatoire Imperial de Paris. 12. A. F. A.*. . 1816. . Annales de I'Observatoire Imperial de Paris. B. and in Journal du genie civil 11/2 (1846). Annales de I'Observatoire Imperial de Paris. table V (repeated in his Traite des fonctions ettiptiques et des integrates EuUrietmes. Tables trigotiometriques detimales. 3. Bigourdan. Paris. no. calculator extraordinary'. in response to a ministerial request. 17. E. 13. Comptes rendus de {'Academic des Sciences. 7). precedees des tables des logarithms. but there are Blagden materials elsewhere. Proceedings of the Royal Society of Edinburgh. 8 (1872-5). 4 (1858). 624—9 (also as Chambres des pairs. Lefort. 4 (1858). Sang. E. [146]. P.'Description des grandes tables'. impres- sions diverses. 6 (1993). de Prony himself reported that during the run of the project. 45 (Summer 2002). G. E Lefort. [123]-[150]. vol. Proceedings of the Royal Society of Edinburgh. 18. J. This letter is quoted in P.Additions. C.. Moniteur universel (1840). "Remarks on the. Lefort. Paris. M. See C. Annuaire du Bureau des Longitudes (1928). 2. F. 1801. 'Reply to M. [146]. E. 1826). These pages are in the file in the Archives Rationales cited in footnote 4. F. [123]-[150]. [123]—[150].J. Riche de Prony. 1800—1830: a neg- lected figure in the history of French mathematics and science*. 8 (1872—5). 994—9. Legendre. 581—7. Exerdces de cakul integral. under the direction of M. and also in his 'Description des grandes tables'. Lefort. This was a huge sum of money. 260-8. F. Notice sur les grandes tables. completed by J. . pp. Prony*. This information comes from the obituary C. tables computed. Courcier. [146]. and the prime numbers thereafter up to 10000. P. [123]-[150]. 'Eloge de M. 11. 421—36. Dupin. 999.'Note sur les deux exemplaires manu- scrits des grandes tables'. See the table in I. Additions. Proceedings of the Royal Society of Edinburgh. Craik. A1-A72. especially A25-A28.. 'The ingenieur savant.. 'Description des grandes tables'. 4 (1858). A. 563-81. vol.. P. Annales de I'Observatoire Imperial de Paris. 19. Additions.. . 'Observations relatives aux remarques.. In fact. 46 (1858). For a recent survey ot Sang's logarithmic work see A. 'Le Bureau des Longitudes'. 20. A.338). Borda.. Sang. 405-33. 15. 'Edward Sang (1805-1890). Huzard-Courcier.The let- ter is not among the Blagden papers in the archives of the Royal Society. 4 (1858).. about five times a typical annual academic income of 13000-18000 francs. Delanibre.120 THE C O M P U T A T I O N F A C T O R Y 10. 48. 32-43. he had his Bureau prepare a set of some trigonometric tables to nine decimal places (G. 16. He took the logarithms for all odd numbers from 1163 to 1501. BSHM Newsletter.'Description des grandes tables'. Paris. p.'. M. Paris. but he gives no provenance. 8 (1872—5). Didot. C. le Baron de Prony*. Sdetxe in context.

pp. 1886. ... V... Firniin Didot. et de sinus et tangentes. 22. Nouvelles tables astronomiques et hydrographiques.. 1891. Paris. Plan and Nourrit. 21. Paris. Bagay. THE C O M P U T A T I O N F A C T O R Y 121 This set is in the archives of the Ecole des Pouts et Chaussees along with various other manuscript tables: see the Catalogue des manuserits de la Bibliotheque de I'cole des Fonts et Chaussees. Imprimerie Nationale. 4—5. Service Geographique de FArmee. Paris. 1829. Tables des logarithmes.

45 .

Tomash.I The Swedish difference engine builders Georg and Edvard Scheutz published their Specimens of tables calculated. stereomoulded.) . 5 Difference engines: from Muller to Comrie MICHAEL R. and printed by machinery in London. but many are unaware of several others that were produced after Babbage"s efforts and even one mention of S. WILLIAMS Most people in the computing disciplines know of the famous difference engine designed by Charles Babbage. 1857. it was perhaps the first set of tables to be published that were entirely machine produced. (Collection of E.

Johann HelfKch Miiller (1746-1830) was a military engineer who later was employed by the Governor in Giessen.W. giving very sketchy descriptions of his ideas for improved devices. but some were of use in the actual pro- duction of extensive tables and. This was a modest four-function calculating machine based on a design first created by G. evidently written by Miiller himself. Difference methods were once of fundamental importance in the con- struction of tables but they have fallen into disuse since the invention of the digital computer. This device was typical in that it had mechanical difficulties that prevented it from being easily used.1 This work consists of 47 pages of instruction for the simple machine and three pages. mainly a printing mechanism for a calculating machine. it is worth a small digression (see sidebar. p.124 DIFFERENCE ENGINES the concept that predates him. he made his own version. When he heard about Hahn's machine and its difficulties. they were already of lesser importance during the era of the massive electro-mechanical machines developed by. 125) to explain these methods before examining the machines that used them. Almost nothing is known of Mullet's design (if indeed it was anything other than just the concept) but he indicates that if someone would be . No one can claim that difference engines were a mainstay of table-making ventures. Germany. Because of their relative obscurity. if nothing else. Bell Laboratories and Harvard University in the 1940s. but he certainly could not be said to have been in the commercial production of them. One of the very early machines (perhaps the first that could be said to work even reasonably reliably) was produced by the German clergyman Philipp Matthaus Hahn in 1774. Indeed. A small portion of Mullet's appendix is devoted to an entirely different machine which. as such. he claims. they make an interesting story about the efforts to mechanize the drudgery of the task. edited and published a user's manual in 1786. he designed to produce sequences of calculations by the method of differences. among others. were prone to problems which were overcome once they began to be produced on a commercial scale in the 1800s. Leibniz over a hundred years earlier. Philipp Engel Klipstein. Mutler's machine Many of the early mechanical calculating machines were constructed on a one-off basis and. It apparently worked well enough that a friend. Hahn and his son made several different versions of this device.

and these may be evaluated by difference methods for short intervals before a new approximating polynomial must be used. or greater orders of differences. We have thus replaced the two multiplications and two additions required to evaluate the function F by two simple additions. To obtain the value . On the other hand it becomes a daunting task if you require all values of x between 1 and 100000. In the very early days of mechanical calculating machine design. A"). the contents of each order of difference to the one above it and finally adding the contents of A to the register containing F(. For example.\-2 + 2x + 3 = 6 11 18 27 38 51 66 83 102 First difference. In terms of modern technology.. Every polynomial whose largest power is x" will have a constant »rth differ- ence and thus is amenable to being calculated in this way. the construction of such a machine is trivial. then the effort is quite manageable if you are only interested in values of x between I and 20. it will be necessary to take the differences between these differences (the second differences: A~ and possibly even the third (A").The machine must be capable of adding.F(10) it is not necessary to evaluate the original fimction: one need only add the constant second difference (2) to the last first difference (19) to get the fact that F(9) differs fiom F(tO) by 2 + 19 = 21 and thus F(10) must be F(9) -f A 1 = 102 4.21 = 123. a tabula- tion of the quadratic fimction rioted above would yield: x=* I 2 3 4 5 6 7 8 9 F(x) = . The method of differences involves tabulating a few values of the function and then noting the differences between these values (A'}... It is unfortunate that this method is only applicable to the evaluation of polynomials but non-polynomial functions {such as logarithms and trigonometric functions) may be approximated by poly- nomials. DIFFERENCE ENGINES 125 Difference methods and difference engines If one Is faced with the task of producing a table of the values of a polynomial function. and several others to hold the various orders of differences (A1. The replacement of many multiplications by a simple sequence of additions results in a huge reduc- tion in the labour of table making. . it was both a major conceptual and physical task. 5 7 9 11 13 15 17 19 Second difference. 2 2 2 2 2 2 2 Obviously the second difference is a constant value 2. A 2 .. This is true even if relatively few values are required but die function is quite complex. fourth (A ). say x~ •+• 2x + 3. A 'difference engine' is simply a calculating machine that contains several registers: one to hold the value of F(x). For functions more complex than a simple quadratic. in sequence.v) to generate the next function value.

5 . my head leaning forward on the table in a kind of dreamy mood. The origins of Babbage*s ideas are not at all clear. and seeing me half asleep. of course. Other than the fact that the machine would be capable of dealing with three orders of differences (which would have been required for a cubic function). what are you dreaming about?" to which I replied. that Charles Babbage (who had an extensive collection of books on ail aspects of calcula- tion3) was evidently unaware of its existence4 until his friend John.126 DIFFERENCE ENGINES willing to finance the construction he was willing to make one. It is not entirely clear whether nobody was actually interested in such a device or the limited circulation of Klipstein's book did not bring it to the attention of the people in need. It would appear that no one answered Muller's plea for financial aid and that nothing carne of this idea until Babbage independently thought of it about 35 years later. Today Klipstein's book is quite rare2 which would indicate that it was not widely circulated in his own day. Herschel gave him a copy that he had found during a trip to the Continent. with a table of log- arithms laying open before me.'Well Babbage. no other details are mentioned. very likely using the printing mechanism he had in mind for the simple calculator noted above. coming into the room. Another member. the most famous difference engines. even if one only worked for 8 hours per day. for example. that sometime in 1812 or 1813: One evening I was sitting in the rooms of the Analytical Society.The following will simply touch on a few highlights of the whole story. He estimated that a table of x3 (I ^ x ^ 100 000) could be pro- duced in just over 10 days of effort. let alone the full facts that led up to them. It is quite impossible in a short chapter to describe these devices in detail. The Babbage machines The machines designed by Charles Babbage (1791—1871) are. He indicates. but that it was certainly possible to create additional copies by employing an ordinary workman to turn the handle of the machine. so readers are advised to consult several of the books and papers that deal with this subject in depth discussed in the Further Reading. It is notable. 'I am thinking that all these tables (pointing to the logarithms) might be calculated by machinery'. at Cambridge. in his autobiography. He claimed that the machine would print a single copy of the table being calculated. called out.

particularly when the members of the Royal Society praised Babbage's efforts. creating superb accuracy and finish on pieces that did not actually require it. capable of evaluating any quadratic to six figures of accuracy. the best mechanic in London at the time. it was certainly spurred on by the experience he had publishing his famous set of logarithms. and the material from which it should be fashioned. As Babbage also possessed some of these same attributes. Babbage had hired a mechanic. describing it as: ".9 This step was both the salvation and ruin of the Engine construction project..producing figures at the rate of 44 per minute. long careful consideration of the design of each piece. of course. The Government later added to this amount and kept adding to it. After asking the Royal Society to comment. Clement was. perhaps. We know that he was working on the prepara- tion of his logarithm tables and simultaneously considering the problems involved in the construction of the difference engine in the early 1820s. had the opposite approach. He had shown it to a number of friends and wrote to Sir Humphry Davy. . Clement was a perfectionist and was capable of the finest workmanship— a talent that was badly needed by Babbage at this stage of his ambitious project. to help him with the con- struction. By 1822 he had constructed a working model of a difference engine. required further reports from the Royal Society. although he had only recently arrived there from the North of England. the President of the Royal Society. the tools to make it. DIFFERENCE ENGINES 127 This was written 50 years after the reported incident and Babbage admits that he did not remember the occasion but a friend (likely John Herschel) related it to him. This led to long periods when the Government would not automatically pay Babbage's accounts. rather than it should be done quickly. Babbage had received /1500 with which to begin the work. and performing with rapidity and precision all these calculations. It is also clear from Babbage's own writings that he was heavily influenced by the work of de Prony..7 He took great pains to ensure these were correct and it must have been obvious to him that other tables could benefit from some form of mechanism that would ensure the accuracy of their calculation.'8 The purpose of this letter was to get the Royal Society's support for an application to the Government to pay for the construction of a full-sized machine.6 Whatever the true origin of his inter- est in table making machines. Unfortunately Clement's attitudes also extended to the creation of the most exquisite drawings. The Government. and generally . by July 1823. the two of them were more concerned that it should be done well. Joseph Clement. the Government agreed and..

in 1833. . as a demonstration piece for the Government and this device now resides in the Science Museum in London.128 DIFFERENCE ENGINES delayed the progress. caused the two men to part company and the construction of the machine to come to a standstill. over whether Clement should move his workshop into a new fire-proof building on the grounds of Babbage *s house. Thus Babbage found himself in the difficult position of being unable to continue with the construction even if he had managed to hire another mechanic.2 Demonstration piece of the difference engine. 1833. Finally. When Babbage and Clement parted company. Clement exercised his rights under British law and kept all the took and drawings that had been made for the difference engine. The only parts that were even assem- bled were. after many years. but by that fig. Clement eventually returned the drawings. 5. a dispute.

A3. mainly because of the bad experiences in attempting the construction of the difference engine. Behind this first row was another set of axles for the adding mechanisms. This was the only folly complete set of plans that Babbage ever produced—all his others were for unfinished versions of his engines. These axles also contained the carry mechanism.11 He had no intention of ever attempting its construction but produced them to satisfy his own ideas of what he owed to the Government. A4. The front would have shown 7 vertical axles (one for each of the 6 orders of differences and the 7th for the function value) each of which would carry 18 brass wheels engraved with the digits 0-9 to represent the numbers. A2. a friend and promoter of science (although not a good scientist him- self) wrote an article about the difference engine for the Edinburgh Review}2 . A3. he took time off from his studies of the analytical engine to create a set of drawings for his Second difference engine. the carries resulting from these additions were dealt with. it would have been about 10 feet high. in 1849. A* were added to A1. 10 feet wide. the numbers stored on A'. and behind that was the equipment to engage and dis- engage the adding devices when required. 4. 3. the numbers stored on A2. and a crater on the Moon. His name appears on several geographic features: a mountain range in Western Australia. The Scheutz machines When Babbage was having difficult times with the Government. A5 were added to F(x). Had the difference engine ever been finished. and 5 feet deep. Dr Dionysius Lardner. 2. the resulting carries were taken care of.10 However Babbage was mind- ful of his previous arrangement with the Government and. a river in the Yukon Territory of northern Canada. A4. Of course Babbage was justly famous for more than his difference engine and he is especially remembered for his analytical engine and his work in economics. The Engine was to operate in 4 distinct cycles. A5. A printing mechanism would be used to stamp the function value into a soft substance which could then be used to make stereomoulds for print- ing the tables. DIFFERENCE ENGINES 129 time Babbage had conceived of the analytical engine and. each corresponding to a quarter turn of the drive wheel: 1. vowed never again to venture down that road.

Lardner took liberties in the description and changed some of the physical aspects of the machine. machine to what he thought was Babbage's design—but wasn't because of the liberties Lardner's had taken in his explanation. they received a small grant enabling them. with professional help. Lardner's article happened to be read by a printer from Stockholm. gown* situation where. and submitted it to the Royal Swedish Academy of Sciences for approval. One of the major ones was that he described the number storage registers as consisting of a hori- zontal row of rotating rings. It was eventually purchased for use in the newly formed Dudley Observatory in New York State. the town won and the research oriented astronomers left for more congenial situations. aided him in this work. Although Babbage supplied the information. By 1843 they had completed a machine. After staying in London for a time while the Scheutzs looked for a buyer. It was capable of working to four orders of difference. to finish their full-scale dif- ference engine in 1854. in this case.130 DIFFERENCE ENGINES This was mainly done to promote the idea and urge the Government to continue with the financing arrangement. but it also provides us with quite a good view of what Babbage had in mind for the design. but by 1837 they began to make metal parts and could see that a complete machine was feasible. although used to produce a few . with only two 5-digit registers and the printing mech- anism. When being worked by an experienced operator it could generate 120 lines of a table per hour. after some difficulty. Their initial experiments were in wood. it generated the logarithms of numbers from 1 to 10000 in just 80 hours. This vertical arrangement resulted in less fric- tion and thus less power being used to effect an addition. His son Edvard. The Scheutz Engine was brought to London where Babbage gave it his unreserved praise. In one experiment done by the Scheutz team. The Scheutz difference engine. Babbage had actually decided on the vertical arrangement of the numbers. The story of the Dudley Observatory is a classic 'town vs. including the time taken to reset the differences for the 20 different approximating polynomials. Scheutz produced a technical journal and thus made a point of reading the latest technical articles from other countries. each number being kept to 15 digits. it was taken to Paris for the Great Exhibition and was awarded their Gold Medal. He was immediately struck with the elegance of the concept and began experimenting to construct his own. who was a student at the Royal Technological Institute. The Scheutz father and son then asked the Government for a grant to create a full-scale machine and. Pehr Georg Scheutz.

being hooked up to a windmill).3 Drawing of the Scheutz engine In Dudley Observatory. set about creating his own difference engine. It now resides in the National Museum of American History. . after 64 years.This second copy of the Engine was the one machine (other than the work of Comrie to be described later) that can be said to have made a major contribution to the creation of mathematical tables. from Report of the Astronomer in Charge of the Dudley Observatory for the year 1863. Martin Wiberg (1826—1905) was a well educated Swede (Ph. like the Scheutz team before him. was purchased by Dorr E. which may have helped in securing some official financing. at one point. limited calculations (even. In the 1850s the British Registrar General decided that it was time to pro- duce a new set of life tables for use in the growing insurance industry and it was decided that a difference engine should be obtained to aid in their pro- duction.D. The story of the use of this machine by William. asked the Swedish government for money. Felt (the inventor of the Comptometer) for his collection of calculating devices. knowing of the Scheutz machine. Farr is described by Edward Higgs in Chapter 8. Wiberg machine Another difference engine was created as a direct result of the work of the Scheutz team. It would appear that his motive was not to produce these machines for sale. in Washington DC. 5. He is known to have experimented with several mod- els and. was largely ignored and. He appears to have had a champion in Prince (later to be King) Oscar. DIFFERENCE ENGINES 131 Fig. but to use them to create tables for a scientific press. Smithsonian Institution. from Lund University) who.

entered the Lawrence Scientific School at Harvard College. in 1869. involving an attractive face of type. however. appear to have been a successful publication and today they are very rare. however. Wiberg certainly attempted to produce other tables using this device.'16 Wiberg exhibited copies of the logarithm tables in the Centennial Exhibition in Philadelphia in 1876.132 DIFFERENCE ENGINES Fig.13 They noted that a set of interest tables14 had already been calculated and proposed that the Academy give its approval to this beauti- ful and ingenious machine. some 20 years earlier. and of different sizes.) A Wiberg machine was sent for exhibit at the Great Exhibition in London in 1861 but arrived too late to be shown. It is not entirely clear if these problems were associated with financing or mechanical difficulties but it resulted in a 13 year delay in their appearance in print. attended Dartmouth College and. 5. exam- ined by Babbage and he wrote letters of reference for Wiberg to introduce him to French scientists which resulted in the machine being examined by three members of the French Academy of Sciences who published their report in 1863. He was born in Maine. is certainly a great advance over the product of the Scheutz machines. He is known to have calculated a set of logarithm tables that could have been published shortly after his Paris visit but were delayed while he worked out problems with the printing. with top and bottom tails. Because of his previous .4 The Wiberg difference engine. Grant's difference engine George Bernard Grant (1849—1917) was the son of a ship builder. Tomash. It was.15 As noted by R. (Collection of E. C. Archibald: 'The result. They do not.

another description indicated that it was not a completely fin- ished product. square and cube roots. which was exhibited at the 1876 Centennial Exhibition in Philadelphia.. The combination of the several parts is extremely simple. Prof. and the machine acts with the greatest certainty.20 After graduation he continued to experiment with designs of calculating machines and took out patents relating to them in 1872 and 1873. Benjamin Peirce.. the number of elements can be indefinitely increased. Perhaps it is not surprising that it was not fully functional considering that it had such a short gestation from its conception and was. pos- sesses great merit for the simplicity and certainty of its action and for the ease with which its parts can be adapted to the requirements of any given function. and give it up again'. After spending some time contemplating the difficult mechanical movements it would require. He was unaware of the work of either Babbage or Scheutz but knew of the existence of simple mechanical cal- culators. only finished a few days prior to the opening of the Exhibition: It occupies a space of about five feet in height by eight feet in length. as through the liberality of the superintendent of the Coast Survey. During his first year at Harvard he had to laboriously calculate a 'cut and fill* table and this prompted his first thoughts about building a machine to do the same job. About 1870 he learned of Babbage and his difference engine and this inspired him to return to his own and he 'designed a machine that might possibly have worked. evidently. While the above quotations certainly imply that the machine worked properly. reciprocals. containing. mainly Wolcott Gibbs. but I could convince nobody that it would do so. Once again Wolcott Gibbs came to his aid and arranged financing for the construction of a full-size machine. sines and tangents and their logarithms. when in full working order. DIFFERENCE ENGINES 133 experience at Dartmouth. the efficiency of the design for its purpose may be considered as having been proved. he gave up the study of such a machine and returned to manual methods of table construction. etc. which had worked to satisfaction'. he was able to complete the normal four-year program in only three years and still have time to think about the con- struction of calculating machines. and weighs about two thousand pounds. from twelve .. such as logarithm. The report of the Exhibition Commissioners described it as follows:21 This machine for computing and tabulating mathematical tables.18 and John Bachelder.17 At the urging of some of his professors.19 he again returned to the design and'Though I have built no large machine. I have been able to build a model of small capacity. is arranged to combine and print functions involving one hundred elements.

1876.. Most of the exhibits were given to the Smithsonian Institution (and formed the basis of their museum collections) but they have no record of ever having received Grants machine.134 DIFFERENCE ENGINES thousand to fifteen thousand pieces. The machine is driven either by hand. by a crank at the front end. 5.22 Grant's Engine was. According to one biographer: 'Shortly after graduation. and in the case of the present machine—the first ever constructed of so complex a character—imperfections are to be expected which will not exist in future machines. while at the front end is an apparatus for printing a wax mould of the results. to be donated to the University of Pennsylvania when the Exhibition was over... or by a power appliance at the rear end. Thirty of the elements of this machine were placed in a light -wooden frame and worked successfully at a speed of over one hundred terms per minute. All that limits the speed is the imper- fection of the mechanism. It is quite likely that Grant simply took it back with the intention of putting it into proper order. this speed would be perfectly predicable.. as a result of his calculating-machine work. a speed may be made of ten to twelve terms per minute.The most reasonable explanation for its disappearance is likely contained in the 'imperfections' hinted to in the above quotations. provided that the mecha- nism of the driving-gear and printing apparatus were in accurate working order and made sufficiently strong to stand the wear and tear resulting from the same.5 Grant's difference engine on exhibit at Philadelphia. he started a machine-shop for Fig. In fact there appears to be some mystery as to its eventual fate. and from twenty to thirty when by power by the attachment at the rear end. . The long body contains the calculating mechanism. When the machine is worked by hand. from The masterpieces of the Centennial International Exhibition. under the conditions of the financing arrangements. Neither can it be traced to the University of Pennsylvania. and if the whole machine were used and sufficient power applied.

There is. The same biographer indicated that he continued his interest in the area and 'during the last years of his life he conducted considerable exper- imental work in connection with the development of such machines'. Julius Bauschinger (1860-1934) and Johann Theodor Peters (1889—1941) began a project to construct a new set of 8-place logarithm tables of both the natural numbers and the trigonometric functions. however. they asked Christel Hamann about the possibility of constructing a difference engine to help with the work. Hamann machine By the beginning of the twentieth century the technology of mechanical calculating machine was already well advanced with commercial versions of several different types readily available.24 Simultaneously with the start of hand calculation. no hint as to whether this work was concerned with dif- ference engines or simply with the four-function mechanical calculators he was also known to have designed and sold. Johann Peters was a professor in the Royal Astronomical Calculating Institute in Berlin and an expert on methods of table produc- tion. particularly the so-called 'proportional lever' which was used in the Mercedes Euklid automatic multiplication machines. The resulting machine is illustrated in the frontispiece of the two-volume set of logarithm tables25 and a detailed . Most of the well-regarded tables of the day only contained 7 places of decimals and these were found to be inadequate because of the recent advances in astronomic observations and geodesy. DIFFERENCE ENGINES 135 gear cutting. The German engineer Christel Hamann (1870-1948) was in the forefront of this technology and had invented a number of devices.. At the time of this project Julius Bauschinger had just left his position as Director of the Astronomical Research Institute in Berlin and had become Director of the Imperial Observatory at Strasbourg.. He was thus an important figure in astronomy and appreciated the usefulness of logarithms to all branches of science. and became one of the founders of the gear-cutting indus- try in the United States. At about the same time. Archibald describes Peters as 'perhaps the greatest mathematical table maker of all time'.'23 His efforts at founding a business likely caused him to put the difference engine aside and he simply never got around to cleaning up the final details.

.27 Many other devices were designed to help people with calculations using the method of differences. Several versions of the famous Brunsviga calculators were produced that were specifically designed for difference methods. Among them was one built by Alfred Deacon of London. the numbers being each of 20 digits. The resulting 2 volumes of tables were published in 1910 and 1911. it was only necessary to use the machine to interpo- late between short intervals. When used by an experienced operator. in consequence of its author having read Dr. including the time taken to enter the initial values. It appears that Deacon read the same article by Lardner that inspired Georg Scheutz. In order to ensure 8-digit accuracy in the results.136 DIFFERENCE ENGINES description of its internal workings can be found in the first volume's pref- ace. the machine could both manipulate and print 16-digit numbers. He began to experiment with various mechanical devices and eventually pro- duced a small difference engine that would operate with three orders of differences. the calculation of 36 table entries would take only 5 minutes. In one.26 It is entirely possible that Charles Babbage later acquired this machine (which is now apparently lost) because he offered the Great Exhibition of 1862 '. The machine consisted of two inde- pendent arithmetical mechanisms and a simple printer. the Brunsviga Dupla.. made in London. a task which could be mainly accomplished using only the second differences. Because of the way Bauschinger and Peters had set up the work. Hamann delivered the machine in 1909 and it was immediately put to work. The Hamann difference engine was a great deal simpler than any of the previously described devices. there were two result registers with the ability to transfer their numerical contents back on to the setting levers—such a machine could be easily used to calculate tables from their first or second differences..a small difference engine. Other minor difference engines A number of other difference engines are known to have been constructed. Lardner's article in the Edinburgh Review". It was not automatic in its operation but required the user to enter initial values and then repeatedly turn the two cranks to add the second difference to the first and then the first difference to the function value. Unfortunately the machine itself has been lost. Perhaps the most ambitious of these devices was .

At the Nautical Almanac Office Comrie quickly developed a reputation as an expert in all aspects of mechanical calculation and his advice was often sought on the best uses of various new machines. It was in 1931 that he examined the National Accounting Machine Class 3000 (produced by the National Cash Register Co. usually take it apart. Different registers were activated (by metal clips attached to the printing carriage) as the carriage moved various columns of figures (debits. one machine could be transferred mechanically to the setting levers of the machine below. in 1917. . credits. Born in New Zealand. Thompson who mounted four standard calculating machines in a wooden platform and modified them so that the result from. His acquaintanceship with massive calculation problems developed into a life-long passion when he became a graduate student in Cambridge. DIFFERENCE ENGINES 137 constructed by A. as part of the New Zealand Expeditionary Force to fight in the French trenches of the First World War. He would obtain copies of each new machine as it appeared. but also sold under the name Ellis) and noted that it contained a number of extra registers which were designed to be used for the standard accounting sub-total applications. Leslie John Comrie (1893-1950) and the National Accounting Machine Leslie Comrie was one of the great figures in table making of the twentieth century. he attended Auckland University and graduated with a degree in chemistry in 1916. etc. by the Fall of that same year he had enrolled in University College London where he was first introduced to machine calculation by using a Brunsviga (a pin-wheel type machine of the original Odhner design). He used this aid when calculating a new set of 20-place logarithm tables28 published in the early 1950s. and then produce a description of which jobs were best suited for its use.29 After a short spell of teaching in the USA. in February 1918. Remarkably. but they could be used as the main accumula- tors for the device. he returned to Britain to work in the Nautical Almanac Office. The interesting thing about these extra registers was that they were not simply used for temporarily storing results. After graduation he joined the army (even though he was very deaf) and was sent. His military service came to an abrupt end when.) in the account sheet over the printing mechanism. he had his left leg blown off by a British shell. J.

While recognizing that tables of all kinds would still be of use for some years. In 1936 Comrie left the Nautical Almanac Office over a dispute about the use of the equipment and staff for non-governmental table making. 5. . and founded the world's first computing service bureau. Comrie checking a set of tables. was highly sceptical of the use of large com- puting machines for the production of tables. A great many errors in earlier tables were found from Comrie's use of the National as a difference engine and were reported in Mathematical tables and other aids to computation and elsewhere. which had a very significant influence on the British computing scene. While Comrie certainly used this device in making some of the many tables he produced in both the Nautical Almanac Office and in his later private ventures. he thought that the pro- duction of tables by Aiken's Harvard Mark I was simply a waste of effort. He suffered a series of strokes in late 1949 and 1950 and died in December 1950 at age 57. legendary table maker and champion of the use of mechanical equipment. Scientific Computing Service. its main help was in using difference methods to check older tables for accuracy. Comrie realized that this automatic selection of which register was active at any given moment could be put to good use to make the machine into a difference engine operating with six orders of difference.138 DIFFERENCE ENGINES Fig.6 L ). It is interesting that Comrie. The different registers could also be brought into a calculation by depress- ing keys on the left of the machine. In recent times a crater on the far side of the Moon and an asteroid have been named in his honour.

it is always best to read the original sources where possible. although it must be treated with care. P. has been added. again to Babbage's design. Only one of these machines—the second copy of the Scheutz machine employed by Farr—was heavily used. Clement. as stated in the introduction. 1864 (numerous reprinting*) contain references to his engines. This was created from the design left by Babbage in 1851 but constructed using modern manufacturing methods. However. on the 200th anniversary of Babbage's birth. in practical situations. then the balance sheet looks a little better. DIFFERENCE ENGINES 139 Were the difference engines important? There is. and his workmen. Of course when one takes into account the other manufacturing advances that came from the efforts of Babbage. certainly helps to show that Babbage's ideas were rather more than what was necessary at the time. It has been used to tabulate complex polynomials and. too. the London Science Museum unveiled its construction of Babbage's second difference engine. Recently a printing unit. gives credence to the idea that Babbage's designs were more advanced than was required by the scientific community of the day. Another interesting source is the book by Babbage's son H. with great difficulty. I think. That. Every attempt was made to use materials that might have been available to Babbage and to ensure that the individual parts were no more accurate that Joseph Clement could have managed in the mid-1800s. particularly his autobiography Passages from the life of a philosopher. Further reading When attempting to find further information on any difference engine. of course. Babbage. one other difference engine that must be mentioned. Babbage's calculating . Many of Babbage's contemporaries have been vilified because of their statements that the money spent on his devices was a waste of resources. The fact that several different working difference engines were created in Babbage's lifetime (and shortly after) using technology that was a great deal simpler (and smaller) than that employed by him. In 1991. is certainly up to the task. The result certainly proves that Babbage's design works. that the marginal use made of difference engines for subsequent tables helps to confirm that these comments were not all misguided. The works by Charles Babbage are often available in large libraries and many of his writings. no one can claim that difference engines were a major factor In the production of mathematical tables.

and Co. Bexhreiburg seiner neu erfundenen rechenmaschine. MIT Press. Glory and failure: the difference engines ofjohann Mutter. For a discussion on this point see M. 1990. Mass. Cambridge. 1989). The complete story of the Scheutz difference engines can be found in Michael Lindgren's Glory and failure. . Frankfurt. 3. Charles Babbage. we are fortunate that the best account of the work of Comrie is contained in Mary Croarken's fine book. This book not only gives lots of detail on Babbage and his machine.. E.. Cambridge. Little. and Georg and Edvard Scheutz. 1889. so one must rely on the work of historians who have had the privilege of researching these topics in depth. 235-40. MIT Press. Oxford University Press. 1. Perhaps the best modern work on Babbage's machine is Doron Swade's The cogwheel brain.Williams. in the works noted in the references.'The scientific library of Charles Babbage'. both analog and digital. Notes and references In the notes below 'WB' refers to Works of Babbage. She not only details the life and accomplishments of Comrie (including a description of how the National 3000 Accounting Machine was actually used). but does an excel- lent job of discussing ail forms of computing.140 DIFFERENCE ENGINES machines. In 2003 a copy was offered for sale at a price of $20. but covers the difference engines created by Miiller and Babbage as well. Brown. Annals of the History of Computing 3 (1981). 1990. but chronicles the interesting story of the Museum's efforts to produce the first of Babbage's large-scale machines. This highly readable work not only deals with the Scheutz machine. London. 4.. The work of Martin Weiberg is similarly scattered throughout short items. Pickering & Chatto. 2 from the original plans. up to the creation of the digital computer in both Manchester and Cambridge. M. 1990. There is little mate- rial easily available on the difference engines of George Grant other than his original papers noted in the references. London. Early scientific com- puting in Britain. edited by Martin Campbell-Kelly (11 vols. 1786. Mass. 2. 2000. R. Lindgren. often containing no more than a brief mention. Unfortunately the majority of the creators of difference engines did not feel it was worth documenting them in ways that would be easily available to the modern reader. Klepstein. Swade was the curator at London Science Museum which constructed Babbage's differ- ence engine No. On the other hand.000.

1860. DIFFERENCE ENGINES 141 5. 19. was the Rumford Professor and Lecturer on the Application of Science to the Useful Arts from 1863 to 1. 1876. Passages from the life of a philosopher.113-17. G. R. 7. 135-43. London. L. M. Corapagnie anonyme de Forsete. 30-31. II. Grant. WB. New York. C. third series. Mathematical table makers. The Edinburgh review 59 (1834). Actually his third. J.. 2 vokjWilhelm Engelmann.Archibald. London. C. Dictionary of American Biography.. J. 14. 'Babbage's calculating engines'. 118-86. M. E. 23. Leipzig. Reports and Awards. 27-32. 20. London.VIII. Annals of the history of computing 9 (1987).'On a New Difference Engine*. 24.. Longman and Co. 330-9.Williams. Mathematical tables and other aids to computation 2 (1947). R. M. The masterpieces of the Centennial International Exhibition illustrated . Logarithmisch-trigonometruche tafetn tnit acht dezinmlstellen enllmltend die logarithmen alter zahlen von 1 bis 200000 und die log- arithmen der trigonotnetrischenfunktionenfurjede sexagesitnahekunde des quadmnten. 3. known as the 'greatest chemist in America* during the nineteenth century. 1876. Babbage 'On the division of mental labour". Spoil and Co. Wolcott Gibbs. his tables and difference engine*. 1910-11. B. 15. 8. Volume 7.'Babbage and Bowditch: a trans-Atlantic connection'. WB. Babbage. 283-90. 'Charles Babbage's table of logarithms (1827)*.191-202. 263-327. 'Joseph Clement: the first computer engineer'. R. M. vol. 18. C. Wiberg. D. 159-69.. John N. 69-76. R. 1889. 42 WB. Bauschinger and J. . Locke. P. 17.'On a new difference engine'. C. 9. Economy of machinery and manufactures. Archibald. Stockholm. T. 25.887. 56 (1863). August 1871. Bachelder had been the man in charge of the Scheutz machine at the Dudley Observatory and was thus one of the few who had practical experi- ence with the operation of such a machine. p. L. Babbage. Babbage's calculating machines. Comptes rendus. 1864. 12. M. 13. US Centennial Exhibition. 22. Stockholm. Philadelphia. 1948.'MartinWiberg.Wilson. Grant. 16. Lardner. Med Maskin utraknade och stereotyperade Rante-tabeUer. 1876.Williams. Wiberg. Scripta Mathematica. Annals of the history of computing 1C) (1988). if you count his initial trial model from 1823 as being the first. pp. 10.es et imprimees au moyen de la machine a calculer. H. entry for George Bernard Grant. 371—73. Campbell-Kelly. See for example. XI. Peters. M. 1835. Tables de logarithmes calcule. Annals of the history of computing 14 (1992). 21. 6. Jemte en Dagriiknings-Tabell. 11. American journal of science.

H. Specimens of tables calculated..Thompson. 28. 29. 194—222. Babbage's calculating machines. stereomoulded and printed by machinery. 197.924. 1952. P. WB. and E. II. London. Cambridge. Scheutz. 1857. G. A.J. 27. Babbage. . Logarithmica Britannia. privately published.142 DIFFERENCE ENGINES 26.D. Comrie was awarded a Ph. in astronomy in 1. 2 vols.

This page intentionally left blank .

paper drums in raised position taking an inked impression from the printing wheels. 2 printing mechanism. 2.1 (b) Difference Engine No. showing inking rollers and ink bath (upper).1 (a) Difference Engine No. Fig.Fig 6. Argument is left-hand three figures. experimental printout for a seventh-order polynomial. . 6.

certainty. The originating agency of literary texts is irreducibly human and. The mechanization of other processes.1 In the second half of the century electronic technologies brought new levels of speed. proof reading. economics. In the electronic era of the twentieth century the printed table as a computational aid. The need for reliable computation and permanent printed record was met by computers and calculators with on- line printers. largely. The theme of this chapter is the interplay between machines and the wider processes involved in table making. 6 The 'unerring certainty of mechanical agency': machines and table making in the nineteenth century DORON SWADE When machines are mentioned in the context of table making it is their role as calculators that tends to be emphasized. Paradoxically the technology that conclusively solved problems of reliability in the production of printed tables all but eliminated their need. and distribution.The dis- tinction is not a pedantic one. and portable devices provide calculation at the place and time of need. specifically transcription. and convenience. portable or desk-bound. typesetting. while tables can be viewed simply as numerical . printing. have received less attention perhaps because they are seen to be part of book production and not distinctively belonging to table making. vanished. and technology of their production is shared with the bibliography of literary texts. As material artifacts volumes of printed mathematical tables fall into the category of'books* and as such the history. though not entirely. During the first half of the cen- tury electromechanical punched card equipment and electrically driven mechanical calculating devices were developed and pressed into service. One clear respect in which printed tables differ from literary texts is that their con- tent (apart from prefaces) is essentially numerical not alphabetical. binding. The successful transition from manual to machine-produced tables is largely a twentieth-century phenomenon.

Those of the eighteenth century. they differ from literary texts in at least one fundamental respect in that the numerical content of many tables can be generated by machines that embody mathematical or computational rules. In this and other late . but were expensive. and correctly lifting and repositioning the moveable carriage for each digit of the multiplier. the correct physical operation of the machine. The execution of single arithmetical operations such as the multiplication of two numbers using an arithmometer is not entirely automatic but relies on the continu- ous informed intervention of the operator. They worked. marketed under various names of which Brunsviga is amongst the best known. and few were made. and Muller.3 The correctness of the final result therefore relies not only on the repeated correct mechanical functioning of the device but on the faultless execution by an operator of a sequence of manual procedures which surround and intervene in the process. Stanhope. objets de salon. Even reliable manual calculating devices have limited use in table making. The device often billed as the first to be made commercially for general sale is the arithmometer introduced by Thomas de Colmar in 1320. notably by Halm.146 UNERRIN6 CERTAINTY OF M E C H A N I C A L AGENCY texts. it took over half a century for de Colmar*s calculator to emerge as a product robust enough for daily use and it was not until the latter decades of the cen- tury that they sold in large numbers. unre- liable and unsuited to daily use.2 Because of their unreliability alone arithmometers had practically no impact on mathematical table making until the last decades of the nineteenth century. Harms and Mullet's in particular. The seventeenth century devices were ornate curiosities. rotating a handle the correct number of times for each decade. as do those of the eighteenth century. It was not only the time it took for such devices to evolve from erratic novelties into reliable workhorses that inhibited their use. It is per- haps because of this distinction that the starting point and ultimate emphasis in discussions about technology and table making tend to centre on calculation at the expense of processes generic to book production. By the time the arithmometer had matured as a product. and Leibniz in the seventeenth century feature routinely in histories of computation. Pascal. the correct input of data. Multiplying two numbers involves entering the digits on sliding dials. desktop machines. However. are extravagantly ornamented testaments to the instru- ment makers' art. The mechanical calculating aids devised by Schickard. So there is a dimension to the role of machines in the bibliography of table making that is distinct- ive and unshared by literary text: the mechanization of authorship. and the error-free transcription of results. came on stream.

and for their cumulative summation. print- ing. reduced in the number of significant digits. Perhaps more importantly. The machines were fast. The use of mechanical calculators is further circumscribed by the place of computation in the overall process of table making. The first patented cold-metal composing machine for typesetting was by William Church in 1822. extended. like all its cold-metal successors. Recomputation of existing tables was exceptional and computation ab initio was only resorted to when new tables were needed. In the nineteenth century table production relied on the same well-established trades and practices as literary book production i.e. So while the ability to machine- generate calculable tables is a distinctive feature of certain classes of numerical tables. These were rechecked. reprinted. Because of the exact- ing labour of calculation and checking. typesetting. held ordin- ary type in magazines above the machine.This machine. merged. Instead. and distribution. com- puting tables from scratch was avoided whenever possible. Those that did used them prim- arily for periodicals and newspapers where speed was more important than the niceties of typography. for example. this feature is not a defining one for the genre. and these are released one at a time by an operator using a keyboard. and anxieties about accuracy. but justi- fication (the alignment of text to produce straight margins at the edges of the page) was still manual. Tables of observational data. Printing and stereotyping Until the 1970s the primary medium of transmission and dissemination of the informational content of tables was print and paper. binding. reputable existing tables were used as the starting point. Cost savings on labour were not great and few nineteenth-century printers used them.4 . In the case of the arithmometer this amounts to the calculation of products of pairs of numbers once entered. proof reading. often owe little or nothing to calculation as an originating process and represent a class of table that invokes many generic aspects of book production. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 147 nineteenth century examples the supposed infallibility of mechanism provides at best security against error only for part of the overall process. not all tables originate in the calculation of mathematical functions. The success of early attempts to automate manual typesetting were neither immediate nor complete. revised. or otherwise modified for new editions to meet new needs.

resetting type broke the line of integrity with the original edition.8 . While the fundamental physical process of inked impression on paper as an inscriptional record remained unchanged. as important a person as the author in the matter of tables. Flong is laminated paper made from alternate layers of blotting paper and tissue paper bonded together. dried and hardened by heating in situ and the mould so formed was then used to cast metal stereotype plates for use in conventional printing presses. Stereotypes rep- resented an immutable form of information capture that offered immunity from the inherent vulnerability of moveable type to derangement during printing or storage. During this time the efficiency of printing presses increased dramatically as machine presses replaced manual presses. whose annotated bibliographies of tables are a rich source of contemporary attitudes and practice. Particular editions of tables acquired their cachet for reliability over decades and reputable volumes of early tables remained in demand sometimes for centuries. The manufacture and economics of paper production changed radically over the century. was rarely retained in its set form and it was common practice to redistribute type for reuse after the edition was out. because the printer.5 The expense of paper as a proportion of overall book production costs dropped significantly. throughput increased and labour costs dropped. By 1900 more than 99% of paper was machine made. Augustus de Morgan. The use of'flong' which came into general use in England in the 1850s ran in parallel with plaster-mould stereotyping.7 Stereotyping had particular attractions for table makers.6 Major influences here were the sourcing of rags used as the raw material for paper. Dampened flong was beaten onto the face of the type. by hand.148 UNERRIN6 CERTAINTY OF M E C H A N I C A L AGENCY The process most directly affected by machinery was printing and paper making. Perhaps the most significant development for table makers was the prac- tice of stereotyping which involved making printing plates (stereotypes) from casts taken from the type already set. Plaster-mould stereotyping was used in literary book production from the 1780s and continued in use till the early twentieth century. However. and the effect of automated paper making machinery. who is. observes that a second edition derives no authority from the goodness of the first. In 1800 all paper was made. expensively. has again stepped between the latter and the public. as already observed. and reset type from scratch for later editions of the same work. Type. the costliest part of a printer's equipment. the abolition of excise duties. and prices had dropped tenfold.

and deferred the costs of replacement raising yields on capital. Charles Babbage commented that errors occurred 'in works where neither care nor expense were spared' but he exculpates the table makers.9 For a combination of economic and information management reasons stereotyping became the preferred technique for reducing the risk of corruption in the press and keeping standard books in print. in storage for future editions increased trade for type founders. prolonged its useful life. verification. The benefits of stereotyping to printers were less clear cut and the intro- duction of the technique suffered the prejudices of compositors and type founders protective of their trades.10 Before turning to one of the most ambitious attempts to mechanize tables production it may be useful to review the separate manual processes . Continuous redeployment of type spread wear more uniformly across the stock. For table makers stereotyping became a hallmark of accuracy and in bibliographies of tables. The appeal of machinery was the promise of certainty. Errors were an unavoidable fact of life and vigilance alone was insufficient to prevent them. For the printers. and proofreading—continued to rely on the mental and physical labour of people. and tying up expensive type. So strongly did de Morgan feel about the benefits of stereotyping tables that he recommended enforcement of its use.. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 149 Storing the stereotypes provided an economical way of preserving the investment in typesetting and proof reading. if an edition was the product of stereotyping. A more subtle benefit to the printer of the reuse of type was the distribution of wear. once set. He wrote that it was 'but just to the eminent men who presided over the preparation of these works. to observe that the real fault lay not in them but in the nature of things'. Calculating and transcribing Humans are fallible and the risk of error was a central and continuing concern for table makers. committing expensive type to storage represented unused capital and few books were kept set up in type as a result. and extended confidence in highly reputed tables to subsequent editions. Stereotyping released type stock for reuse in new work. Resetting type for a new edition from scratch was welcome new work. But the essential informational processes—calculation.. this was noted as an implied recommendation. typesetting. transcrip- tion.

Repetitive calculation was a hideous drudgery and the prospect of using mechanical aids to relieve the worst of its burdens features as a motivational drive in some of the earliest attempts to use machines. In the production of mathematical tables. Subtabulation runs of one hundred to two hundred values between pivotal values were not uncommon. calculation was typically divided between at least two distinct groups. Most major non- polynomial functions can be approximated in a sufficiently small domain by a polynomial expansion and this allows them to be tabulated using finite differences.12 In cases where subtabulation using finite differences was used the mathematicians also calculated the pivotal values and the set of starting differences for the computers. and close enough to ensure that the interpol- ated values stayed true to the requisite number of end digits.11 The advantage of the method is that it reduces the arithmetical process to simple repeated addition and avoids the need for multiplication and division that would ordinarily be needed to evaluate the function. The contrast between 'excellent men' and 'peasants' or 'slaves* also reflects a class difference which pre-echoes Babbage's description of de Prony's three groups amongst whom the labour was divided. Pivotal values were chosen suffi- ciently far apart to reduce their overall number (and so transfer the greater burden onto the computers). mathematicians and computers. Instead of substituting incre- mentally increasing values of the argument into the mathematical formula and redoing the calculation to produce each new tabular value of the function. as 'classes' with the computers referred to as . interpolation techniques were used to find intermediate values between more widely spaced pivotal values. In 1685 Leibniz wrote that 'it is unworthy of excellent men to lose hours like slaves in the labour of calculation which could be done by any peasant with the aid of a machine'.13 The reference to 'excellent men* and 'worthiness' implies a hierarchy of values in which abstract. One benefit of the technique is that it deskills the calculation process and allows the repetitive work to be carried out by computers with only rudimentary arithmetical skills. analytical and philosophical activity is ranked higher than repetitive task-specific activity. Mathematicians chose the formulae.150 UNERRIN6 CERTAINTY OF M E C H A N I C A L AGENCY involved and their respective vulnerabilities to error. fixed the range of the table (the start and end values of the argu- ment) and decided the decimal precision. A common technique for such sub-tabulation used the method of finite differences in which each next tabular value is generated by a series of addi- tions from a precalculated set of starting differences.

mathematicians and scientists were more than happy to offload routine calculation onto the computers. and one set of logarithms to nineteen places.'that wearisomeness and disgust. 'the intolerable labour and fatiguing monotony of a continued repetition of similar arithmetical calculations'.16 In the absence of machines. 'the dull and tedious repetition of many thousand consecutive additions and subtractions'. One reason for this is to ensure that rounding errors did not accumulate during the successive additions to the point at which they affect the least signific- ant digit. "And what discouragement'. The effectiveness of the technique relies on the independence of the . at the time an engineer. which always attend to monotonous repetition of arithmetical operations'. wrote of the stultifying effect of such drudgery on higher thought.17 The generic precision of tables for some applications was in any event high. Checking With the calculations complete manuscript results were checked for errors. If each had done their work perfectly the separate sets of results would be identical.'does the perspective of a long and arid computation cast in the mind of a man of genius'. and verification took the form of consistency checks performed on the separate outcomes. De Prony's cadastral table of natural sines was tabulated to twenty five decimal places. the 'mental drudgery* of constructing tables. The inclusion of additional digits to ensure correctness to large numbers of places com- pounded the volume of routine numerical calculation as well as increased the burden on the mathematicians who were responsible for specifying the number of such 'guard digits' sufficient to ensure the integrity of the end digits in the printed version.14 The numbing grind of routine calculation is described with vividness and conviction in contemporary writing. he asks. A standard technique was for the same set of calculations to be performed by different computers without collaboration. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L AGENCY 151 the 'third class' performing the 'lowest processes of arithmetic' and ordinal- ity providing a scale of superiority. If fifteen significant digits were needed in the answer the calcu- lation would often be carried out to twenty or twenty five decimals.1* Luigi Menabrea. who attended Babbage's seminar in Turin on the Analytical Engine in 1840. The burden of calculation was further increased by the need to work to a larger number of decimal places than required for eventual printing.

they were found copying from each other. This applies to manual methods as well as subtabulation by automatic machines. If subtabulation is continued one value beyond the end of the run then the new tabular value should be the same as the next pivotal value. produced the same incorrect result.23 The two independently computed sets of de Prony's grand Tables du Cadastre survive in manuscript and a visual comparison gives an indication of the frequency and nature of corrections made to reconcile the two inde- pendently calculated sets or results. moralizes with characteristic puff that 'falsehood in this case assumes that character of consistency.152 UNERRIN6 CERTAINTY OF MECHANICAL AGENCY computers from each other. hired to assist in the preparation of the British Nautical Almanac. Such occurrences would elude detection using coincidence checks. by reversing the process of subtabulation by repetitive addition (numerical integration by differences) and subtracting successive tabular values already calculated. a prolific popularizer of science. despite insulation from each other.18 The technique of double computation was not foolproof and it was not unknown for computers who. to return him the same numerical result. The first observation to make is that the number of corrections is very large and it is clear from this that the .22 The pivotal values provide a further check. Finally. In 1770 two computers were instantly dis- missed when.20 To further reduce the risk of error the separate computers were sometimes given com- putationally different but mathematically identical formulae to find the same result. there is verification by differencing i. and that result wrong'. Dionysius Lardner. Small errors produce wild fluctuations in the higher order differences. Since each value is used in the calculation of the next tabular value it follows that the last value depends on all prior values. The task of verifica- tion can therefore be reduced to a single final-value check. The technique of subtabulation using differences has self-verification features which assisted error detection and lightened the burden of verifica- tion. working separately and independently.e. which is regarded as the exclusive attribute of truth*.21 The intention here is to disrupt shared habitual mental patterns or common psychoperceptual tendencies originating this class of error. Lower order differences can sometimes be checked by visual inspection and mental arithmetic and any deviations signal that something is amiss.19 He reports that on occasions de Prony found 'three and even a greater number of computers. The technique is very sensitive. Since the pivotal values are calculated inde- pendently the coincidence of the two values provides confirmation of cor- rectness.

Typesetting The corrected and transcribed manuscript results were then given to the printer where the first process was to set the results by hand using move- able type. the profusion of corrections confirms the wisdom of dupli- cating the calculations and of verifying by consistency checking as insisted on by the table makers. Overall. their distribution does not follow any evident pattern. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 153 contribution of errors from the combined processes of calculation and transcription was substantial. or systematic transcription errors from correctly calculated val- ues. There are many corrections to single entries by overwriting and these are episodic i. A common form of this would be say to increment all the third digits in a column of entries by'!' down the whole page.24 Without the computers' worksheets it is impossible to tell whether these are systematic calculation errors from the method of differences correctly transcribed. By far the largest number of corrections are to sys- tematic errors. and blocks of numbers to make up the page. Selection depends on the position of the box in the matrix and in general an individual piece of type is not inspected as it is used. numbers have no immediate meaning. and the compositor has no intuitive sense of whether one digit has any sensible relationship to the one before or after. It would seem that typesetting numbers rather than language text would pose special difficulties in that while groups of letters make up recognizable words. Typical of this category is the overwriting of one digit in the same position for all entries on the same page. Compositors retrieved individual digits from boxes filled with samples of the same character and the correct selec- tion of type is guided by the habitual action based on the location of the box in the matrix of boxes.e. Surprisingly. typesetting does not appear to have been a source of great anxiety to table makers. In an extreme case a run of forty seven consecutive double-page openings has either two or three columns of each page over- written in one digit position incrementing the original digit by 'I'. Augustus de Morgan comments on . Here the compositor reads from the authored source and sets each digit of each result in loose type to form the groups. lines. Errors can arise by retrieval of characters from the wrong box (mental lapse) or where the wrong species of type is distributed in the boxes before retrieval ('foul case').

once cast.26 Proof reading The first stage of printing was the production of proof sheets for checking against the manuscript. ordinary pains being taken with the previous proofs'.25 It is clear that the high standards of accuracy for literary texts extended to numerical tables. One hazard was the 'drawing of type* which occurs when the ink is too thick or the type not secure and the adhesion between the type and the inking roller or leather inking balls draws the type from the frame. A further reason why checking stereotype proofs attracted superordinate priority was that any errors found could still be corrected in moveable type. It was not unknown for type to simply fall out and be replaced incorrectly. De Morgan urges that 'the strictest investigation should take place in the proof which is taken from the stereotype. which caused die absence in question. the most rig- orous controls are reserved for checking the stereotyped proofs rather than for the proofs taken from. whereas the stereotype. or a temporary suspension of our own quickness of perception. was amenable to only minor correction.27 It is curious that while meticulous care is urged for each stage of the process. a reputable compositor will improve the integrity of a manuscript source and professes surprise at the accuracy with which 'first-rate London printers can turn out their proofs. Indeed the standards of top printers appear to have been disconcertingly high. results set in moveable type. the integrity of the printed result depends on the moveable type remaining entirely undisturbed during the whole printing process and there were clearly anxieties about the vulnerability of type to derangement between the printing of the first proofs and producing the stereotype plates.154 UNERRIN6 CERTAINTY OF M E C H A N I C A L AGENCY how. to the extent that the absence of error led to a sense of insecurity about the quality of error checking. . We have frequently looked at page after page of table-matter more times that we should otherwise have thought necessary. merely because the total absence of detected error left it an unsettled point whether it was the excellence of the proof. Proofreading is an exacting and tedious process and this was reflected in the high pay rates for such work. The reason for this uncharacteristic relaxation of the highest standards of care throughout is that however scrupulous the checking of proofs from type. in the case of literary texts. even where the manuscript is criminally bad'.

the other two to check. reading a result from the manuscript. was the use of "author traps'. Failure by the checker to find the prescribed number of errors per page serves as a warning of faltering attention or fatigue. He urges a modus operandi using multiple listeners where possible. and this is reflected in the variety. is proportional to the distance the eye has to carry the numbers'. without col- laboration. but offers special advice for solo checking in cases where the luxury of multiple listeners is unaffordable. He counsels that the manuscript and the proof copy are brought into close proximity so that the two numbers being compared are contained in the same unbroken field of view. then taking the next reading from the proof and so on. The location of these is concealed from the author-checker but carefully regis- tered by the printer. sub- tlety and ingenuity of the practices and ploys devised to counter fatigue and ensure effectiveness in error detection. Alternating the datum for each reading is advised i. one to read aloud from the manuscript.e. This involved the printer being requested to make. the checker is advised to break patterns to relieve tedium.28 Checking proofs from the stereo- types was the safeguard of last resort. at his discretion.30 Whatever the preference. Another technique used. and he disdains physical wear on the manuscript recommending that it is folded as frequently as every two or three lines 'so as always to have both manuscript and proof under the eye in one position*. and varying the pitch or tone of voice in the spoken repetitions. for example) and against transposition (012 for 102). If either proof or man- uscript is harder to read. De Morgan recommends a checking process involving three people. especially for inexperienced checkers. the hazards of fatigue and tedium . identical printed proof copies from the stereotypes. moving hands and feet. Shifting bodily position. he suggests reading the easier first as 'for the mind is apt to allow knowledge derived from the more easy to give help in inter- preting the more difficult'. a fixed number of deliberate mistakes in every page. making the comparison with the proof. some favour- ing visual checking in silence. The wisdom of this practice is articulated by Glaisher who writes that 'it is well known that the number of errors. are recom- mended best practices. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 155 For any substantial correction the mould of the whole page would need to be redone after correction of the type. Different checkers appear to have had different preferences. and serves as a quality control alarm. others by vocalising the numbers. Clearly...29 De Morgan warns against the tendency to mistake double figures (744 for 774. combin- ing ear and eye.

and bibliographies annotated. Glaisher reports that Vega offered a reward of a ducat for every error found in his table of 1794. The response of the table makers and users to error-ridden tables was to uncork the bottle of red ink and correct them. Pond succeeded Nevil Maskelyne as Astronomer Royal in 1811 and the appointment carried with it responsibility for the annually issued Almanac. reviews written. and many entries are accompanied by a critique of accuracy. Knowing which tables were reliable was an issue of connoisseurship and publishers sought testimonials from eminent math- ematicians to preface new editions. When the unreliability of tables overflowed into scientific and political arenas. For the most part the knowledge of deficiencies in tables was confined to the community of tablemakers and users.3* Rewards were offered for the detection of errors after publication. how many and how gross errors both eye and ear.156 UNERRIN6 CERTAINTY OF M E C H A N I C A L AGENCY were many and real. were near valueless without a record of'pedigree' or 'stable'. insecurities remained. An analysis of the relative frequency and magnitude of last-digit errors published in 1851 by Gauss would have cost Vega dear: there are large numbers of hereditary errors fromVlacq on whose tables Vega's are based. As many errors as possible were corrected prior to printing and the process of progressive improvement continued after publication through the distri- bution of errata sheets and the incorporation of discovered errors into later editions and revisions. will suffer to pass. including errors of a unit in the last digit. how much the perceptions are dulled by the monotonous comparison of one column of figures with another. the problems were seen as failures of superintendence. It turns out that Vega was fortu- nate that no one took up the challenge.32 It is also possible that a few deliberate errors were left as a protection against unauthorized copying. which served as comprehensive consumer guides. poor leadership. or negligent recruitment of appropriate staff—organizational shortcomings in the execution of estab- lished practice. both as an advertisement of confidence in accuracy and as a way of improving subsequent editions through 'user-testing' in the field. Errata sheets would be published. Poor tables were not spared damning reviews. The decline of the Nautical almanac under John Pond is a case in point. when tired. The genealogy of specific editions of tables was an essential determinant in their suitability as a source and the annotated bibliographies of tables. and however ingenious the measures. So battered was the Almanac's credibility by 1818 that it . De Morgan observed: It is hardly credible. to those who have not tried.

The evidence of errata sheets is undeniable but there were those. and errors in calculation blurred issues of . none of the protagonists of mechanical solu- tions cites an instance of a navigational error that resulted in the loss of a ship being attributed to a tabular error.3* Arguing from the premise of known errors was less than conclusive and the inability to defin- itively quantify residual errors left the issue of reliability open to differences in expert opinion. George Biddell Airy included.M Lardner cites a survey of a collection of tables and uses published errata as the barometer of accuracy. who asserted that tables were already sufficiently accurate.The errata were corrected by further errata and his triumph is ecstatic at the finding that the Nautical almanac. upon which it is impossible to say what wrecks may have taken place*. The implication of Lardner's survey is that tables were generically flawed and that the large numbers of errors in the host tables discredited the printed errata sheets as a corrective process being itself vulnerable in the selfsame ways. and the affair was debated in Parliament.3* He compounds this apparently shocking statistic with details of equally egregious errors in tables for astronomical navigation citing seven folio pages of errata to supplementary Almanac tables which themselves con- tained more than HOC) errors. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L AGENCY 157 was seen as an embarrassment to English science especially on the Continent. The unreliability of charts. The problem was not the known errors. Maritime safety was a central and continuing concern and shipwrecks featured routinely in newspaper reports which were often luridly vivid. doubts about observational procedures. captures the anxiety of the unknown: 'an undetected error in a logarithmic table is like a sunken rock at sea yet undiscovered. but the unknown ones. Henry Goulburn in 1842. Just how bad were the tables? A frequently quoted source is Lardner's lengthy semi-popular article on Babbage's engine published in 1834. writing to the Chancellor. Babbage observes that'the multitude of errors really occurring is comparatively little known'. of 1836 would require an erratum of the erratum of the errata. He reports that in a random selection of forty volumes of arithmetical and trigonometric tables (nautical and astronom- ical tables were excluded) selected from a collection of 140 volumes over 3700 acknowledged errors were found as evidenced in the errata sheets.37 It is curious that though astronomical navigation features prominently in the advocacy for the need for accurate tables. the Astronomer Royal.33 The scandal focussed on issues of organization and leadership. John Herschel. the deficiencies of which resulted in lowered standards and higher error rates.

158 UNERRIN6 CERTAINTY OF MECHANICAL AGENCY accountability. Most British ships lost at sea foundered on the coasts in storms. c. (Science Museum/Science & Society Picture Library) .2). Experts disagreed whether Fig. 1832. While errors in calculation for tables were the first stimulus for his work it is an oversimplification to see the engine designs as a 'solution' to a widely recognized and clearly defined 'problem'. Dismayed at the number of errors Babbage famously exclaimed 'I wish to God these calculations had been executed by steam'. 6.'Steam' was a metonymic reference to the 'untiring action and unerring certainty of mechanical agency'39 and he devoted the major part of the rest of his working life to the realization of this mechan- ical epiphany.The genesis occurred in 1821 when he and Herschel were checking manuscript calculations for astronomical tables commissioned by the newly founded Astronomical Society.38 Mechanized table production The most ambitious attempts to mechanize tables production in the nine- teenth century were those of Charles Babbage whose efforts are increasingly well known (Fig. 6.2 Charles Babbage (1791-1871).

It is known that Babbage col- laborated with Lardner on the article and it is easy to assume that Lardner was Babbage's mouthpiece—that the views Lardner expressed were Babbage's. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 159 the deficiencies in tables were serious and whether or not the engines would be of any use. and the technical and intellectual challenges of this pro- gramme soon consumed his interests. are prominent in the earliest expressions promoting the utility of his machines. and a way of promoting Babbage's damaged interests at the same time. In developing his engines Babbage was led from mechanized arithmetic and tabulation to fully fledged general-purpose programmable automatic computation. Babbage's calculating engine represented fresh material for his lecture tours and writing from which he derived substan- tial fees. . Lardner's own agenda is also relevant. However. Eliminating errors in tables provided the entry point to a new discourse of computation and as the work progressed tabular errors were subsumed under the larger ambitions of general-purpose computation. the timing and content of Lardner's article needs to be examined in the political context of the collapsing fortunes of the construction project which after a decade of engineering design and mas- sive public expense had little in the way of credible progress to show. the stimulus to new forms of analysis. The cost benefits of building engines for table making com- pared to manual methods were disputed at the highest level40 and the ambiguous successes of Scheutz and Wiberg support these reservations. 1 and this has had a defining influence in the historiography of Babbage's work. the heuristic value of the machine in the discovery of new mathematical series. Here the parade of known tabular errors is paired with the unabashed promotion. perhaps elaborated and dramatized for the purposes of public consumption. and the saving of mental and phys- ical labour. Tabular errors feature alongside other arguments in his advocacy. Table making remained a preoccupation though not always a central one. the use of computational method as an alternative to analytical solution to equations. Creating a 'scandal* of tabular errors may have been a way of simplifying the complex motivational uses of the machines into a pro- blem—solution pairing—the act of a publicist with an astute sense of what was needed for public appeal. He enjoyed a lucrative career as author and public lecturer. of Babbage's engine. Benefits to mathematics. The clearest coupling of tabular errors as 'problem' with machines as 'solution' is found in Lardner's article of 1834 on Babbage's Difference Engine No.

page borders. for the most part unbuilt in his lifetime. Jamming is a form of error detection and Babbage boasts that the machine will calculate correctly or break. 2 which he designed between late 1846 and 1849.The calculating section was completed for the bicentenary in 1991 of Babbage's birth. leaving blank lines between groups of lines. all of which sur- vive. Program options are selected before the . In the study of the designs it is the calculating mech- anisms that have commanded the lion's share of attention. or. and the printing and stereotyping apparatus was com- pleted in March 2QG2. some- where between integer values for any given digit position. Typesetting is automatic and the layout of results on the stereotyped page can be programmed by the operator. and the extent to which such a device might have been. but never deceive.41 Yet integral to Babbage's conception of table-making machines is not only error-free cal- culation but the extension of the supposed infallibility of machinery to the task of printing results. they represent the most elab- orate and ambitious attempt to mechanize tables production. two. and the detail and physical scale of his undertaking provide clear insights into the size and difficulties of the task. the availability of a working engine gives clues about how the machine might have been operated and managed. were offered to the Government in 1852 through Lord Rosse who acted as an intermediary. number of lines on a page. Formatting features that can be varied include line height. However. Security against error is provided by a number of devices and mech- anisms. but the offer was rebuffed and no further attempt was made in Babbage's lifetime to construct the machine. Plans for the new engine.43 The construction of the machine gives insights into the engineering task that faced Babbage. across the page (column-to-column) and automatically line wrap to the start of the next line. practically useful. Results can be stereotyped down the page (line-to-line) and automatically column wrap to the top of the next column. or four columns. three. margin widths.42 The Engine has since been built to the original drawings at the Science Museum in London. The Engine is a decimal digital machine: it recognizes only integral decimal whole numbers as individual digits in a number and the machine is designed to jam in the event that any figure wheel. print wheel or stereo- typing wheel punch takes up a position that is indeterminate i. Moreover.160 U N E R R I N 6 C E R T A I N T Y OF M E C H A N I C A L A G E N C Y Babbage's engines. An elegant illustration of the scope and scale of Babbage's intentions is his Difference Engine No.e. and options to stereotype in one. had negligible influence on table-making practice.

The printer and stereotyping apparatus is also part of the control system of the whole tabulation process. A feature of the machine is that the calculating section can be split horizontally to isolate upper and lower sections of the machine. Programmability is provided by 'pattern wheels*. From the point at which the handle is first turned the internal state of the machine has altered from the initial condi- tions and is unknown to the operator. This is done by inhibiting the . the start state of the engine for the next page will be incorrect and the first result stereotyped in the fresh tray will be spoiled.44 The construction of the printwheels for the inked impression is such that leading zeros can- not be suppressed without dismantling a section of the apparatus. Soaked cardboard and plaster of Paris are being experimented with (Fig. With the Engine halted the operator has the opportunity to replenish the stereotype trays for the start of a new page. two of the materials used by Babbage in contemporary trials of other machines. Suppressing leading zeros in the stereotyped record by replacing the wheel punches with blank spacers is possible but not practical. this is of little significance. One set of pattern wheels controls the line-to-line operation and the 'fact- ory issue' version of the machine comes with four different pattern wheels with variations of tooth pitch for different line spacing. If the full thirty digits are used in the printed result then generating the argument would require a separate calculation run using only one difference to increment each next value.45 Finally. If the operator overruns beyond the end of a stereotyped page. and this was clearly not intended. and the separation of the teeth determines the amount the tray advances to receive each next result. These are flat wheels with teeth cut into the circumference.The end-of-page condition is detected by the stereo- type apparatus and automatically disengages the drive handle. There is an ingenious device that prevents this happening. There are several practical and operational issues that have become evid- ent during early trials of the apparatus. The printing and stereotyping appar- atus is at the opposite end of the machine and the furthest away from the operator. The handle suddenly runs free and the Engine is halted automatically in exactly the state it needs to be to start a new page.3). and it is likely that leading zeros would have been cut or machined off the printing plate after it was cast. The force with which the stereo- type punches are lowered has been found insufficient to make any meaningful impression on lead or copper. 6. there is the question of how tabulation of the argument would have been carried out. Since the printed record is a one-off and intended as a checking record only. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 161 start of a run.

3 Difference Engine Ho. 2. Its complexity and physical size testify to the difficulty of bringing the 'unerring certainty* of machinery to the task of producing error-free tables. Figures are indented into the surface. and the mechanism to do this is simple and easy to operate. In operation it is an arresting sight. polynomial function of up to the seventh order. and a sep- arate section can tabulate the argument using only one difference to pro- vide the linear increment of the argument.162 UNERRIN6 CERTAINTY OF MECHANICAL AGENCY Fig. The argument and result will be printed and stereotyped on the same line alongside each other. The machine in its entirety is a sumptuous piece of engineering sculp- ture. Experimental stereotype mould. 6. It follows that one section of the machine can perform the tabulation of a. The use of the machine to generate the argument and the result in the same run by distributing the full digit capacity of the machine between the argument and result provides a speculative answer to the question of why Babbage designed his machines to operate with so many digits. (Science Museum/Science & Society Picture Library) carriage of tens at any digit position in the thirty-one digits on each of the eight columns of figure wheels. Piaster of Paris. . Separate calculations can then be carried out in each of the isolated sections with the thirty digits apportioned between the two sec- tions at will.

U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 163 Babbage's Difference Engine No. and one that is most evocative of its overall shape. During each cycle the machine executes seven thirty-digit decimal additions to produce each next value in the table which appears on the column containing the tabular value. 6. turning. and sliding motions required to execute the repeated additions required for the method of finite differences.* The drawing shows a machine eleven feet long and seven feet high with the depth varying between eighteen inches and four feet. and impresses the same results on soft material to produce stereotype moulds from which printing plates can be made. The built machine weighs five tonnes and consists of 8000 parts equally divided between die printing and stereotyping apparatus. at the same time. For lower order polynomials the higher order difference columns are set to zero and pky no part. produces ten results per minute. via a set of bevel gears. 47 The cam stack orchestrates and harmonizes the lifting. once run in. The Engine automatically calculates and tabulates any seventh order poly- nomial to thirty-one decimal places using the method of finite differences. The design lias an elegant economy: it is about three times more efficient than Difference Engine No. The results are stereotyped in two font sizes. hundreds next above. The Engine. and so on. tens next above. a set of twenty-eight conjugate cams arranged in a vertical stack. Numbers arc stored and operated on using figure wheels engraved with the decimal numbers *0' through'9'. 1 in having three times fewer parts than the earlier design for a similar computing capability (Fig. large and small. For a seventh order polynomial the machine adds the constant difference on the seventh difference column to the sixth difference column alongside. and the calculating section. The Engine is operated by turning a crank handle by hand. 2: overview Difference Engine No. 2 was designed between 1846 and 1849. and the layout of the results in the stereotype trays can be format- ted by programming mechanical options. Initial values are set from a precalculated table by rotating the figure wheels after disabling the security mechanisms. . 6. and so on. and a column for the tabular result.5). The crank handle also drives die printer and stereotype apparatus from a long shaft which runs the length of die underside of the machine. prints the result to thirty places on a paper roll. The handle drives. Calculation and typesetting are completely automatic eliminating the need for manual transcription. The side view of die machine (Kg. The machine has eight columns each with thirty one figure wheels* one column for each of seven differences.4) is one of the main drawings in a set of twenty. A thirty-one-digit number is stored in a col- umn of figure wheels thirty-one high with. the sixth difference to the fifth difference. Negative values are represented by complements. units at the bottom.

Side Elevation. 6. 2. Difference Engine No.4 Design Drawing.Fig. (Science Museum/Science & Society Picture Library) .

2: printing and stereotyping A central feature of Babbage's calculating engine designs is the automatic printing and stereotyping of results. The machine weights 5 tonnes and consists of 8. Completed 2002. The printing and stereotyping apparatus is . Difference Engine No. 2 incorporates an apparatus which prints in ink a one-off copy of results on a print roll as well as producing a stereotype mould for the production of printing plates for use in a conventional printing press. Foreground: Printing and stereotyping apparatus. UNERRING CERTAINTY OF MECHANICAL AGENCY 165 Fig. Rear: calculating section.000 parts.5. 2. 6. Charles Babbage's Difference Engine No. (Science Myseum/Science & Society Picture Library) Difference Engine No.

showing transfer of results from figure wheel column (right) to print wheels and two sets of stereotyping wheel punches. printing apparatus (computer simulation).6(a) Difference Engine No.166 U N E R R I N 6 C E R T A I N T Y OF M E C H A N I C A L A G E N C Y Difference Engine No. 6. to the print wheels. printing and stereotyping apparatus (computer simulation). This hardcopy printout is intended as a Fig. 6. Fig. and at the same time to two sets of punches for stereotyping (Fig. printing and stereotyping is accommodated in the calculating cycle and takes no additional time.Typesetting is automatic and each result is printed and stereotyped during the calculating cycle in which it is generated: there is no buffeting or storage of the result and there is no time overhead to print. via a system of racks. Paper drums (bottom). 2.6). directly coupled to the calculating section and each new tabular value is transferred automatically from the results column. Each cycle leaves an inked impression of a thirty- digit result on the paper print roll which advances automaticaEy to provide- fresh paper for each line (Fig. 6. inking mechanim and print wheels (upper).6(b) Difference Engine No.7). and spindles. Trays below receive impressions. 2. (Science Museum/Science & Society Picture Library) . i.e. 6. pinions. 2: printing and stereotyping cont.

6. Difference Engine No.The platform or carriage supporting the stereotype trays has two horizontal degrees of freedom which allow the trays to advance down the page (line-to-line) or across the page (column-to-column). During each calculating cycle the punches are driven downwards into the tray below and die material is impressed with all thirty digits at the same rime in one action. with hardened steel number punches fixed at the circumference. margin widths. number of lines on a page. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 167 Fig. Each set consists of thirty wheels. (Science Museum/Science & Society Picture Library) record and checking copy only and the line spacing and format is not alterable by the operator (Fig. one with a larger and one with a smaller font.9). number of columns. After each impression the tray advances automatically and repo- sitions itself to receive the next result on a new line or column and the distance between the impressions (the line spacing) is determined by the incremental advance of the tray. 6. printing apparatus (computer simulation).8). Programmable formatting features include line spacing. and raising to return to receive the next result). There are two sets of stereotype punches (Fig.6(c).10). . Paper drum in raised position taking an inked impression from the plntwheels. 6. The wheels making up the punches have two degrees of freedom only: rota- tional (to register each digit of the result). The carriage is driven by falling weights which are rewound automatically at the end of a column or page. one for each digit of the result. Below each set of punches is a bronze tray to take the soft material to receive the impression (Fig. The layout of results can be programmed by the operator.The wheel punches are assembled on the same shaft and are lowered and raised together as a set. 6. and vertical (lowering to impress the result.The production of multiple copies for dis- tribution is achieved through stereotyping. and blank lines between groups of lines. The smaller tray advances by a proportionately smaller amount to give a reduced line spacing in keeping with the smaller font size. 2.

Printing and Stereotyping Apparatus. 2. Plan of inking. 6. Difference Engine No. (Science Museum/Science & Society Picture I .7.fig.

stereotyping wheel punches for large font. 6.9 Difference Engine No. experimental printout of checking copy for tabulation of seventh- order polynomial Argument is left- hand three figures. 2. 6. (Science Museum/Science & Society Picture Library) Rg.8 Difference Engine No. (Science Museum/Science & Society Picture Library) .UNERRING CERTAINTY OF MECHANICAL AGENCY 169 Fig. 2.

(Science Museum/Science & Society Picture Library) Further reading There is no single reference that comprehensively treats the influence of machinery on.J. An excellent work on computational devices and systems is Michael Williams's A history of computing technology. 2. A concise and authoritative account of scientific computing which seamlessly blends calculating devices with organizational and operational practice is Mary Croarken's Early scientific computing in britain. Plaster of Paris. 2002. 1990. table making. Philip Gaskell's A new introduction to bibliography. Inc. 2: printing and stereotyping cant.10 Difference Engine No. is an impressive source on the history of the processes involved in book production."En^ewooA Cliffs.: Prentice-Hall. Oxford: Clarendon. 1972. and material tends to be drawn from the histo- ries of the separate relevant technologies. experimental stereotype mould. of book production. Rg. Press. 1985. N. particularly the history of mechani- cal calculation. Oxford: Oxford University Press. For a detailed economic anlaysis see Alexis Weedon's Victorian publishing: Continuum International Publishing Group. printing. . and papermaking.170 UNERRIN6 CERTAINTY OF M E C H A N I C A L AGENCY Difference Engine No.6.

Uniquely authoritative accounts of the workings of Babbage's engines are by Allan Bromley in various issues of IEEE Annals of the History of Computing. 7. Science Museum. Chapter 5. Gaskell. 1975. A new introduction to bibliography. 5. 228. and George Wilkins. 6. often inadvertent. A. Oxford University Press. London: Bradbury. For detailed analysis see..* Bulletin of the Scientific Instrument Society 52. p. 2000 by Doron Swade. 2. The last third of the book gives an account of the late twentieth-century construction of Babbage's Difference Engine No. 'Making the Arithmometer Count. Weedon. E Gaskell. 5. Notes 1. 1861. no. 'Trends in book production costs' in Victorian publishing: Continuum International Publishing Group. Science Museum. 1972. See. 11 for detailed account of operation. Chapter 3. Calculating machines and instruments. March (1997). pp. 1842. London: Little. London: Charles Knight. 2002. Agnew. 12-21. and the extended version in The English Cyclopaedia: A Dictionary of Universal Information. 4. An uncompromisingly internalist description of this engine is 'Charles Babbage's Difference Engine No. 201-4. 274. London. See Mike Williams. by the same author. S. Brown. 2. Chapter 11. BaxandalJ. .). Oxford. in this volume. See p. edited by Martin Campbell-Kelly. Science Museum Papers in the History ofTechnology No. A recent accessible account of Babbage's life and the evolution of his thinking on calculating engines is The cogwheel brain: Charles Babbage and the quest to build the fast computer. 11 vols. London: William Pickering. 3. The standard primary-source reference for Charles Babbage's published work is The Works of Charles Babbage. p. 496—501. Ibid. Revised and updated by Jane Pugh (ed.'. op tit. Augustus de Morgan's articles in The Penny Cyclopaedia of the Society for the Diffusion of Useful Knowledge. The prefaces and editors' introductions to published tables tend to be a neg- lected source which sometimes offer rich pickings. Technical Description. are rich sources which serve both as annotated bibliographies of tables as well as revealing repositories of information on contemporary practice and attitudes. U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 171 Much of the material on contemporary table making is'in the cracks'. Johnston. 2. 1996. 976—1016. 1989. D.

172 UNERRIN6 CERTAINTY OF M E C H A N I C A L A G E N C Y

8, A. de Morgan,'Table', in The Penny Cyclopaedia of the Society for the Diffusion of
Useful Knowledge, 496-501, Charles Knight, London, 1842, p. 500. For an
extended version of the article see 'Table' in The English Cyclopaedia: A
Dictionary oj Universal Information, 976—1016. Bradbury, Agnew, London, 1861.
9. Gaskell,ojJrif,,p. 201.
10. C. Babbage, Passages fmm the life of a philosopher. Longman, Green, Longman,
Roberts & Green, London, 1864, p. 138.
11. See Michael Williams, Chapter 5, in this volume.
12. See Ivor Grattan-Guinness, Chapter 4, in this volume. Babbage observes that the
'third class' of computers who were capable of no more than addition and sub-
traction made fewer mistakes than the mathematicians higher up in the hierarchy
of skills. See C. Babbage, On the economy of machinery and manufactures, 1st edn,
London: Charles Knight, 1832. Reprinted, (ed, M. Campbell-Kelly), 'lite works of
Charles Babbage, 11 vok, William Pickering, London, 1989, vol. 8, p. 138, hence-
forth 'Works'.
13. E. Martin, The calculating machines (Die Rechensnushinen): their history and (level-
oment, edited by Peggy Aldrich Kidwell and Michael R Williams, Charles
Babbage Institute Reprint Series for the History of Computing, Cambridge, MIT
Press, Mass., 1992, p. 38.
14. For Babbage's reference to 'classes' and 'lowest processes' see, Works, Vol. 8.
pp. 138, 141 respectively.
15. Sources in order: Babbage, 1822, Works, vol. 2 p. 6; Ibid. p. 15; Francis Baily,
1823, Works, vol. 2, p. 45;JurisJudex, 1861, Works, vol. 1, p. 3.
16. Luigi Menabrea (1842), Works, vol. 3, p. 113.
17. An extreme example of this is given by Babbage's son. Calculations for logarithm
tables were taken to twenty decimal places for correctness to seven figures. See,
B. H. Babbage, 'Babbage's Calculating Machine or Difference Engine':Science
and Art Department, 1872, p. 8. Reprinted: Works, vol. 2, p. 232.
18. E. Forbes,'The Foundation and Early Development of the Nautical Almanac.*
Journal of the Institute of Navigation 18, no. 4 (1965): 391-401, p. 394.
19. D. Laidner, 'Babbage's Calculating Engine' Edinburgh Review 59 (1834):
263-327. Reprinted: Works, vol. 3, p. 134.
20. Ibid, Italics original.
21. See Ivor Grattan-Guinness, Chapter 4 in this volume, p. 114.
22. For a clear nineteenth-century statement of this feature, see, C. Babbage.
Passages from the life of a philosopher, Longman, Green, Longman, Roberts &
Green, London, 1864, p. 50.
23. For a worked example see, M.V Wilkes. 'Babbage's Expectations for the
Difference Engine.' Annals of the History of Computing 9, no. 2, 203-5, p. 203.
Checking by differencing was used in the late eighteenth century in the
preparation of the Nautical almanac and differencing summed groups of entries
was used, for example, in the production of eight-figure logarithm based on

U N E R R I N G C E R T A I N T Y OF M E C H A N I C A L A G E N C Y 173

de Prony's manuscripts and published in 1891 for the Geographical Service of
the Army. See, Preface to Tables des logarithms a huit decimates de$ nombres entien
de 1 a 120 000 et des sinus et tangentes de dix secondes en dix secondes d'arc dans le
systeme de la division centesimale du quadrant publiees par ordre dti Ministem de la
Guerre, Impremerie Nationale, Paris, 1891, pp. HI and IV, George Biddell Airy
suggested in 1856 that the major use of automatic calculating engines might
well be for error-checking existing tables by differencing.
24. See, for example, tabular entries for log 14050 to log 18790 in the set at the
Biblioteque de I'Institut.
25. A. de Morgan, 'Table' in The English Cyclopaedia; A Dictionary of Universal
Information, 976—1016. London: A. Bradbury, 1861, p. 1015. This is a variant of
the same claim carried forward from the version of the article published
in 1842.
26. A. de Morgan (1842), op cit, p. 501.
27. Babbage reports in 1822 that the rate for checking proofs was three guineas a
sheet. See, C. Babbage, A letter to Sir Humphrey Davy, Bart,, President of the Royal
Society, on the application of machinery to the purpose of calculating and printing math-
ematical tables. Cradock and Joy, London, 1822. Reprinted: Works,
vol. 2, p. 13, footnote.
28. Minor changes to stereotypes were possible depending on the nature of the
correction. See C. Babbage, On the economy of machinery and manufactures, 4th
edn, Charles Knight, London, 1835. Reprinted: Works, Vol. 8, see p. 53,
Section 94,
29. J, W L.Glaisher,'Report of the Committee of Mathematical Tables.*, in Report of
the Forty-Third Meeting of the British Association for the Advancement of Science, held at
Bradford in September 1873, (ed. J. W. L. Glaisher), John Murray, London, 1874,
p. 135.
30. Later practice questions the effectiveness of reader—listener pairs. For work at
the NAO see, L. J. Cornrie, 'Computing the Nautical Almanac.' Nautical
Magazine July (1933), p. 16.
31. De Morgan (1842), op cit,, p. 500; (1861), op cit., p. 1015.
32. Glaisher (1874), pp. 138-9.
33. M. B. Hall, All. scientists now: The Royal Society in the nineteenth century,
Cambridge University Press, Cambridge. 1984, p. 1,1.
34. D. Lardner, 'Babbage's Calculating Engine* Edinburgh Review 59 (1834),
263-327. Reprinted: Works, vol. 2, pp. 118-86.
35. Ibid., Works, p. 129. The private individual is almost certainly Babbage. See
M. Campbell-Kelly, 'Charles Babbage's Table of Logarithms* Annals of the
History of Computing 10, no. 3 (1988), 159-69,162,
36. Passages, p. 138.
37. Herschel to Goulburn, September 1842. Royal Society Herschel Archive, Box
27, Item 51.

174 U N E R R I N 6 C E R T A I N T Y OF M E C H A N I C A L A G E N C Y

38. R. Lain, Shipwrecks of Great Britain and Ireland, David & Charles, London, 1981.
39. The phrase is Lardner's. See Works,Vol. 2,169.
40. See, D. Swade, The cogu'heel brain: Charles Babbage and the quest to build the first
computer, Little, Brown, London, 2000, pp. 37-8,141-7.
41. The major work on Babbage's technical work is by Allan Bromley. See,
A. G. Bromley,'Charles Babbage's Analytical Engine, 1838* Annals of the History of
Computing 4, no. 3 (1982): 196-217. 'Difference and Analytical Engines* in
Computing before computers, (ed. W. Asprey), Ames: Iowa State University Press,
1990. 'The Evolution of Babbage's Calculating Engines', Annals of the History of
Computing 9, no. 2 (1987), 113—36. 'Babbage's Analytical Engine Plans 28 and
28a—The Programmer's Interface', IEEE Annals of the History of Computing 22,
no. 4 (2000), 5-19.
42. Swade (2000), pp. 174-7.
43. Ibid, See Chapters 12-18 for an account of the construction of the calculat-
ing section of the engine.
44. Babbage records that he used specially prepared thick paper and plaster of Paris
in his stereotyping experiments. See Passages p. 46.
45. This technique was used in the Scheutz prototype completed in 1843. See,
M. Lindgren. Glory and failure: the difference engines of Johann Mullet, Charles
Babbage, and Georg and Edvatd Scheutz, Translated by C. G. McKay, 2nd edn,
MIT Press, Cambridge, Mass., 1990, p. 150.
46. For an index of Babbage's technical work see, A. G. Bromley. The Babbage papers.
in the Science Museum: a cross-referenced list, Science Museum, London, 1991.
47. For a detailed description of the engine see, D. Swade, Charles Babbage's
Difference Engine No. 2: Technical Description, Science Museum Papers in the History
ofTechnology No. 5: Science Museum, 1996.

This page intentionally left blank

318 VENUS, 1855.

7

Table making in astronomy
ARTHUR L, NORBERG

For millennia moving around on the surface of the Earth, especially the
oceans, presented difficulties for mankind. Astronomers, seamen, naval per-
sonnel, explorers, and boatmen overcame some of these difficulties through
the use of one or more of the following aids: position tables of celestial
objects as a function of time, topographic maps, charts of shorelines and
currents, and the various instruments for making accurate measurements of
the position of objects in the heavens and on Earth. These aids appeared
after a great amount of research on technique and laborious observation
and measurement. This was especially true of the position tables of celestial

Fig. 7.1 The American ephemeris and
nautical almanac for the year 1855
was the first issue from the newty
established (American) Nautical
Almanac Office. The table shown pre-
dicts the positions of Venus for that
year. (Courtesy of the University of
Minnesota Libraries.)

178 T A B L E M A K I N G IN A S T R O N O M Y

objects as a function of time used to calculate the longitude and latitude of
a particular position on the surface of the Earth. A significant amount of
calculation is needed to predict celestial positions, which is complicated by
the fact that relative positions of all objects change because of the moving
platform, the Earth or satellite, from which the positions are measured,
Without this motion, the problem would have been straightforward once
the principal of gravitation was developed. All we would have to do is set
up and solve an equation of motion using the principle of gravitation. Alas,
the problem is much more complicated,
Throughout history, astronomers have studied the motions of the bodies
of the solar system. In the second half of the seventeenth century, these
studies took a monumental step forward with Isaac Newton's (1642—1727)
formulation of the laws of motion and the principle of universal gravita-
tion. The precise application of the Newtonian formalism to the solar system
involved calculation of the interactions of no less than eighteen bodies in
the solar system: given the positions and velocities of all these bodies at an
instant of time, and assuming that the only force among them was gravita-
tional, one should be able to develop equations of motion and to predict
the future positions and velocities of any one of the bodies.
While we now accept that the only force is gravitational, this was not
obvious at the beginning of the eighteenth century. At that time,
astronomers had few conceptual problems in formulating the equations of
motion of planetary objects themselves. Each of these equations contained
a term for the basic two-body gravitational interaction (between central
body and satellite), and a term representing the perturbative effects pro-
duced on this motion by one or more other bodies.The development of ana-
lytical mechanics in the eighteenth century opened the way for a complete
treatment of all the gravitational effects produced within the solar system, as
required to predict the motions of these bodies. Gravitational astronomers,
who formulated equations of motion for the interactions of the planets, dis-
covered that these equations could not be integrated in closed form, and
developed approximation techniques for the successful integration of the
equations.
The history of predicting planetary positions using theories developed in
celestial mechanics was a search for a complete unified set of precepts for
use in computing future planetary positions. However, this search was occa-
sionally beset by related difficulties due to inadequacies in astronomers'
planetary observations and understanding of planetary interactions.

Euler altered the equations of motion to include a disturbing function. we examine some highlights in this history. if the orbits of the two planets are not in the same plane. the plane of the Sun's orbit. and the difficulties with observation. the mutual interaction between the two planets alters the velocities of the planets. depending on the relationship of the directions of motion and the instantaneous positions of the two planets relative to each other. T A B L E M A K I N G IN A S T R O N O M Y 179 Therefore. the search. The drama has a cast of epic proportions and its denouement occurred at the beginning of the twentieth century in Paris with a major assembly of astronomers at an international conference. namely. preparation of precepts. hence a change in the inclination of each orbit with respect to the ecliptic. In this essay. described using a series of trigonometric functions to express angle and position changes as a function of time. Next the computer subtracted the resulting equations from . Primary emphasis in this chapter is on the first stage in the process of predicting future planetary positions. in the three-body problem. rather than the following stage of computing predicted positions. Euler and his successors solved the equations of motion expressed in these variables for the heliocentric position coordinates of the primary planet. Solving gravitational equations using perturbation theory Newton and his immediate successors investigated the simple Keplerian two-body problem using a geometric approach. namely. Furthermore. and made some attempts to include a third body. that planet would follow an elliptical orbit with constant orbital elements. If only the Sun and one planet interacted. expressing the position of the body in terms of the position in a Keplerian orbit and the variation of this position due to the perturbations produced on the planet. focusing on only a few of the principal actors and institutions. where a second planet is added. there will also be adjustments in the angle between the two orbits. To account for the perturbing effects of a third body. the problem of seemingly unstablizing interactions. However. In a major advance in the eighteenth century. Leonhard Euler (1707—1783) and his contemporaries approached the three-body problem analytically. with predictable velocity values over the entire orbit. This effect on velocity can be positive or negative. it took almost 200 years after Newton to realize the goal.

The orbital elements had to be . Use of the precepts makes the final calculation of future planetary positions easier. one can calculate positions in orbits. A complete set of precepts and the tables constituted a 'theory* for the planet under investigation. This selection and the tables referred to for calculation run for 110 pages in volume 6 (1894). contained coefficients that were functions of the orbital elements. The expressions for the heliocentric position coordinates gen- erated from the solution of the equations of motion. But the tables of planetary motion involved more than simply an analyt- ical theory. When direct integration of these more complicated equations failed. described how to use these tables. in Astronomical papers for the use of the American ephemeris and nautical almanac. Using the precepts. either single or double entry. the equations for the Kepler ian orbit and obtained a new set of equations in the second derivative.180 T A B L E M A K I N G IN A S T R O N O M Y Rg. Mathematical astronomers constructed tables. These equations contain many repetitive collections of terms. From the equations for the variations.2 Simon Newcomb published precepts for the planet Venus. Euler and his successors resorted to approximation techniques. A set of instructions. the computer needed only to refer to the tables and sum the selected entries for a given case. containing cal- culations of subsets of these collections. After some substitutions of terms to arrange the equa- tions of motion only in terms of the variations. 7. and the precepts for other planets. called precepts. the equations of motion could be integrated to obtain the equations for the variations in terms of time.

by knowing which terms from the theoretical calculation could be associated with which observed inequalities in the motion of the body under consideration. by making any needed adjustments in the values of the orbital elements employed in the calculations of the coefficients of each periodic term. therefore. perhaps a decade. the amount of rotation of the line of nodes of the lunar orbit2 could be calculated from observations of the Moon's latitude above the ref- erence plane. Examining the effectiveness of the predicted positions as time passed provided a test for the precepts. corrected for such interferences as aberration. or theory. the . i. the calculation and correction of the orbital elements to determine their initial values. Any possibility to reduce this calculation—either by truncating series. the number of terms retained increased and the quantity of calculations rose at a rapid rate. the predicted positions were usually effective (though decreasingly so). with each succeeding generation of astronomers. The theories. and this value could be used in the theoretical terms that exhibited such a variation.' The initial values to specify the initial elliptical orbit were found through cal- culations using the same theory. etc. parallax. Much of this was done from hindsight. As the eighteenth century wore on. These observations. However. and it is not surprising that the tables could produce only approximations to the true positions of the body to which they applied. Gravitational astronomers forced the terms to represent these observa- tions faithfully. and the solution of the equations of motion—involved large amounts of calculation. or developing new methods—without negative effects on the accuracy of prediction was quickly employed in the generation of tables of predicted positions. Thereafter. In spite of the retention of more terms. For example. This whole process—the development of the theory of motion of the planet. a certain disunity of the approach to the solution of the problem. of perturbative motion contributed to the lack of long-term effectiveness of the position tables.e. served also as initial conditions in the equations for the position coordinates. and. (effects that were themselves in need of theoretical refinement). these sets of precepts of the eigh- teenth century were not effective producers of ephemerides for long periods. For short periods. Uncertainties existed in the observations. the results were consid- ered a positive test of theory. gravitational astronomers recog- nized that to improve the precision of the tables many more terms in the series employed in the equations of motion would have to be retained. TABLE MAKING IN A S T R O N O M Y 181 derived from precise observations of the planet to which they referred.

7. called elements. c.3(a). A is the perihelion. latitude.-e. The methods described by Lagrange. c) of the orbit. its deviation from a circle. In this sphere. 7. i. depending on the interaction to be examined. the eccentridty. and declina- tion. Laplace. One reference sphere is a simple projection of the Earths latitude (0° to +90°. The line AA' is the fine of apsides. where (1-/2)AA' is the serai- diameter a of the ellipse.is defined as the ratio (2CS/AA'). a) and shape (the eccentricity. arc the two reference angles. 0 to —90°) and longi- tude (0° to 360°) lines against die celestial background. 7. The positions in Fig. i. the farthest position is the aphelion. that is its size.e. the orbits position in space relative to the ecliptic (the inclination. the orientation of die orbit in space. an object's position is located where right ascension.) The place of a planet's orbit is related to the ecliptic. <i. and the position of the planet at some time. The indicators.and the planet's position in the orbit (specified by angles from certain reference points). that is. and others.3{b) . called the ecliptic. The plane of the Sun's orbit. with respect to the Sun's orbit. (The ecliptic is analogous to die equatorial plane (plane of 0° latitude) in the Earth's reference sphere. Referring to Fig.182 T A B L E M A K I N G IN A S T R O N O M Y Orbital elements Astronomers use several reference spheres and planes for locating celestial objects. /). the closet planet P gets to the Sun and A'. defines a second reference sphere. Angle /is the true anomaly of the planet P. equivalent to longitude. used features of the shape of a planet's orbit.3(b) S. of a planetary orbit give information about the size (the semi-major axis. the Fig. t.3(a) Fig. 7. which is related to the Earth's equatorial plane for predicting future positions.

wherein techniques were used to force the theory to fit one body's motion. Over time this line between nodes oscillates. astronomers use as a reference the points. series expressions for the heliocentric position coordinates. contained terms that were functions of the orbital elements of the disturbed and undisturbed planets. where the orbit of the planet intersects the orbit of the Sun. the longitude of perihelion from the ascending node. known as small omega. varies each day. is the main point of reference for citing the planet & position. 7. Angle i represents the inclination of the planet's plane with respect to the Sun s orbital plane. <y. identical to Fig. called nodes. II. T A B L E M A K I N G IN A S T R O N O M Y 183 position of the Sun. The node at which the planet moves above the Sun's orbit. Astronomers calculate the mean value of the motion and use it in many calculations of position. The line NN' is the line of nodes. and A and A' are the positions of the planet's perihelion and aphelion. the ascending node. identical to the positions in Fig. to. labeled N and N'. Expansions of periodic terms in the equations of motion contain all of these elements. The final. Such a difficulty entered the process in the following way. and methods differ depending on which element is chosen as the variable in solving the many equations for a single planet. along with a. precepts and tables. 7. The values of the coef- ficients in these terms were determined using orbital elements calculated from observations of each of the planets involved. N. measured from the first point of Aires. and in most cases the elements of the disturbing planet's orbit were the quantities altered without respect for the values these elements must . with little concern for the motions of the other bodies involved. resulting from integration of the equa- tions of motion in which the perturbing function R was expressed in terms of the orbital elements. The orbital elements in these coefficients were then adjusted to fit the early observa- tions. the ecliptic. In locating the orbit. and / from Fig. the time between crossings of a chosen line of longitude.3(a). /. Before the theory was used to generate an ephenieris. however. the values of the coefficients in the terms of the series expressions in the tables were checked by compar- ing the equations with past observations of the disturbed planet.3{a). AA' is the line of apsides.3(b). 7. e. are the orbital elements. the ecliptic. were constructed in a semi-empirical manner. Angle O identifies the position of the ascending node of the planet's orbit as the planet moves counterclockwise in this figure. that is. we obtain the angle w. and the motion of the planet. Projecting the position of the planet at time / onto the celestial sphere at B.

and any errors in observation—was discrepancies between predicted and observed positions. which errors would then be transferred to the coefficients in the series expressions for the heliocentric coordinates. the rates of change of the elements of its orbit could be found. and later Lagrange. the mass of a planet. which appeared after only a short time. found that from knowledge of the position and velocity of a perturbed planet. . called the method of the variation of arbitrary constants. but the results seemed easier to obtain than by ear- lier methods. as described above. one planet would have several sets of orbital elements.4 Published by Lagrange in preliminary form in 1766 and refined in the 1780s. hence the elements of an instantaneous orbit.184 T A B L E M A K I N G IN A S T R O N O M Y have to satisfy the primary motion of the disturbing planet. they could be used to determine the heliocen- tric position coordinates of the planet. This method involved expressing the heliocentric coordinates in terms of the elements of the orbits in which the planets moved. depending on which planetary theory was being analysed. The effect of these three difficulties—the arbitrary elimination of terms in the series for R. Neptune. in the 1740s. Euler. the alteration of parameters in the final expres- sions for the position coordinates by comparing the expressions with observation. instead of the coordinates. The amount of computation using this method was no less. used as constants in the equations to calculate Uranus* effect on Saturn were often different than those used to calculate its effect on. Instead of performing the above operations for all the equations of motion. for example.3 This was the germ of the method. and the perturbing forces as functions of the elements. had the idea that some of this computation could be reduced. As a result. Once the expressions for the elements were in hand. he treated the equations for the radius of the orbit r and longi- tude in orbit A in the same way but isolated the equation for the latitude and replaced it by two equations that expressed the perturbing forces hi terms of the variation of the line of nodes and the inclination of the planet's orbit. it eventually superseded all the other approaches. The values of certain orbital elements. and that these variations would then be useful in calculating the actual elements at some time t. developed to a fine degree toward the end of the century by Joseph Louis Lagrange (1736—1813).. Euler. One further complication in the way of obtaining precise predictions lay with possible errors in the early observations. making repeated corrections to the tables necessary.

well known empirically by 1760. 7. In the solution of Lagrange's planetary equations by integration of the equa- tions of motion in terms of the orbital elements. the final terms displayed two types of time effects resulting from perturbations produced on a planet. Short-period variations were easy to identify. and with these . however. because they exhibited time effects that were cumulative. He moved to Paris just before the Revolution and was instrumental in rebuilding science in this period. and focused on ail the significant prob- lems in the field. The periods varied in length. A threat to the system The pursuit of a unified set of precepts for computing predicted positions was interrupted in the last third of the eighteenth century by a new problem. with a net effect of zero over the complete cycle. He interacted with all the major mathematical astronomers of the period. but could usually be associated with times of the order of the revolution of the perturbed and/or the perturbing body around the Sun. Some of these inequalities. TABLE MAKING IN ASTRONOMY 185 Fig. The second class came to be known as secular inequal- ities. and it was this group that received considerable attention in the eighteenth century. for the most part. up to 900 years. which seemed to threaten the long-term stability of the planetary system. winning several prizes. One class of terms—the periodic inequalities—affected the perturbed body in a periodic manner. and were.4 Joseph Louis lagrange's scientific studies in mathematics and astronomy occurred primarily during his adult years in his native Turin and later at the Berlin Academy. most of the long-period inequalities were discovered in the eighteenth century. especially Laplace and Euier. possessed much longer peri- ods. However.

trying to isolate the inequalities responsible for the differences between the predicted and observed longitudes of each planet. theory of games of chance. These investigations were guided by the belief that the secular inequalities were really long-period effects.5 Pierre Simon Laplace is seen as one of the most influential scientists of all time. Laplace. stimulated by several early studies by Lagrange. The differences in the longitudes of Jupiter and Saturn Fig. the irregularity in the longi- tude was thought to result from uncertain knowledge about the mean motion effects. and that profound investigations of the secular variations using Newtonian gravitation would reveal how they affected the solar system in a periodic way. notably Pierre Simon Laplace (1749-1827) and Lagrange. certain astronomers. 7. chemistry. began his ana- lyses of the secular variations by returning to the beginning of the solution of the equations of motion and extending the series expansions beyond the terms kept earlier by Euler and Lagrange. launched a theoretical effort to analyse the secular variations. . physics. to the interactions of Jupiter and Saturn. cosmology. such as the founding of the famous Ecole Pofytechnique. the definition of the scientific disciplines. there was concern that if the secular variations did exist the solar system was unstable.5 Since the longitude terms depended primarily on the mean motions of the planets. Laplace contributed to many areas of astronomy. mathematics. As a result of this latter concern. and the establishment of the metric system. He applied this analysis first. in a memoir written in 1773. His career spanned the turbulent years of the French Revolution. Moreover. and causality.186 T A B L E M A K I N G IN A S T R O N O M Y discoveries came a great deal of confusion about the distinction between the secular variations and long-period effects. yet he was influential in institution building.

Lagrange applied the method of arbitrary constants to the irregularities in the mean motions of Jupiter and Saturn. they do. In a report delivered to the Academy in 1773 and published in 1776. the Sun were not subject to secular variations. Lagrange came tantalizingly close to solving the problem of the irregularities in the mean motions of Jupiter and Saturn. thus narrowing the problem to secular variations in the motions of the line of apsides. and the perturbative action of Jupiter and Saturn on each of the four inner planets. And so it continued over the next decade.Venus. T A B L E M A K I N G IN A S T R O N O M Y 187 appeared cumulative. and this lawful regularity confirmed the belief of Laplace and others that one could predict the motions of the planets over long periods of time using Newtonian gravitation. This led to a search for some periodic terms that could account for the irregularities in these longitudes. within the limit of his series expansions.10 the line of nodes.* Thus. He. In this manipulation of the variation of the mean motion for the changes in the other orbital elements. the work of each man stimulated that of the other. he demon- strated that variations in the eccentricities and inclinations would be limited to periodic inequalities. both Laplace and Lagrange began investigating the secu- lar variations to see whether or not a long-period inequality might account for them. Lagrange announced that the secular variations would cause the inclinations of the orbits to oscillate about mean values from which they would deviate by only small amounts. Earth. Laplace found that the variation within these limits would have little effect on the basic configuration of the solar system. which included the final form of his development of the method of arbitrary constants. in 1776. the points at which planets are fur- thest from the Sun. varied. . Lagrange published a memoir. which. and that this result was valid as long as the major bodies moved in the same direction. Laplace analysed the limits over which the numerical values of the eccentricities and aphelia. the system was stable.8 He demonstrated that the mean distances of the plan- etary orbits from. but from an analysis of the variation of the elements by the method of arbitrary constants the variation of the mean motion turned out to be zero. of course.9 He applied the technique to three cases: the interaction of Jupiter and Saturn. that is. and the longitude of the planet. Lagrange punctuated this work of Laplace with memoirs of his own.7 Later. found that the inclinations of the planetary orbits oscillated about mean values. the mutual interactions of the system of Mercury. again. Indeed. Here. A few years later. In the same year. and Mars. In 1774. too.

and Johann K. Theoretical development continued in the years during which Laplace prepared the Mecanique celeste (published in 5 volumes between 1798 and 1825)." Lagrange revived his investigations of the mean distances of the planets from the Sun. Burg (1766—1834) were both developing precepts for tables of the moon. Burckhardt (1773—1825) and Jacques P. In addition. in his Mecanique celeste provided such a synthesis. Simeon Poisson (1781—1840) initiated similar studies in about 1805. in the process. Laplace. these tables constituted a splendid aid to navigation. a skillful calculation would yield a position at sea within 27 miles of the true position. Using predicted positions of the moon based on these tables.e. Laplace included all this work in several of the appendices to the Mecanique.25 minutes of arc of observation.12 With these tables lunar positions could be predicted within 1. Binet (1786-1856) calculated the terms in the disturbing function to the sixth power of the eccentricity and the inclination. Hence. tried to isolate more inequalities so that their respective tables would be more accurate.188 T A B L E M A K I N G IN A S T R O N O M Y The multitude of papers on perturbations was not in a form useful for the practical astronomer intent on constructing tables of the motions of these bodies. The rapid developments in perturbation theory and their application in the eighteenth century required a synthesis before further definition of their quantitative usefulness to solar system astronomy could be achieved. A measure of the difficulty involved in the analysis of the Moon's motion. i. devel- oped in the 1750s. Tables While Lagrange and Laplace were busily preparing the groundwork for a new programme for the development of astronomical tables. and each. Laplace's treatise came into immediate use. by incorporating the work on perturbations in a general discussion of perturbation theory starting from first principles of Newton's mechanics as developed by Euler. the tables and precepts that attracted the most attention were those of the Moon. The best tables for the Moon's motion were those of Tobias Mayer. or the motion of any of the other bodies of the solar . Practical astronomers were more concerned with short-period terms in order to improve the usefulness of tables. Burckhardt and Johann T. a unified set of precepts. and astronomers made rapid advances in planetary theory in the years following. because of their practical value in determining longitude positions on the Earth. M.

and published them as they became available. etc. . as is done today. with corrections. 7 system for that matter. sponsored the tables. Delambre. in order to know the position. and Burckhardt. which increased the computational accuracy. Venus (1810). of a projected series. The success of the tables in this period was based partly on the improvements in theory. but the authors attributed no general credit to him. J. The French Bureau of Longitudes. and Mars (1811). At that time. and great hopes were expressed each time for their continued usefulness. Burckhardt published tables of the Moon in 1812. Bernhard A. in 1806. Even before the last volume of the Mecanique appeared (1825). it is no surprise that the appearance of Laplace's Mecanique. von Zach (1754-1832) in 1804 and Delambre in 1806 published tables of the Sun. with a new programme for practical astronomy. were still in use in the middle of the century in the British and the then new American almanacs. Each set of tables that appeared in the hundred years after 1750 showed improved agreement with existing observations. Since obser- vations were so much better than predictions from theory.'13 Sprinkled into the section 'on the Constroction and use of the Tables' for both sets were references to Laplace's Mtcanique. Alexis Bouvard (1767-1843) and Jean B. The Bureau published Part One. Both men based their tables on observations of James Bradley (1693—1762) and Nevil Maskelyne (1732—1811) of Greenwich. among whose members in 1805 were Lagrange. this contained Delambre*s 'Tables of the Sun' and Burg's 'Tables of the Moon. von Lindenau (1780-1854) presented tables of Mercury (1813). at the behest of Laplace. until one arrived at a final position. stimulated a drive in the beginning of the nineteenth century to produce new tables for the motions of all the major planets of the solar system and a new table for the motion of the Moon. Delambre (1749-1822) were in the process of constructing tables. Franz X. subtract. can be seen by noting that instruments of this period could measure positions with a precision of two seconds of arc. it was necessary to extract information from. several tables and add. Laplace. and partly on an increase in the quality of observations. and Bouvard prepared tables of Jupiter and Saturn in 1808 and of Uranus in 1821. The success of Laplace's main thesis in the early volumes can be judged by the number of new tables that were published in the few years after 1803. Bouvard. Some of these tables. It was not an easy task to compare the resulting ephemeris or predicted position tables with observation. Ephemerides of this period did not furnish geocentric positions for any body on a particular date.

But to this value one did not add the remaining advance of die body to the true time. . sixteen calculations were neces- sary. the angle of inclination of the ecliptic and the celes- tial equator. and using the mean time. up to 16h 7' 19". in 1806. Using these tables and repeating the above calculations. which yielded yearly changes in the mean lon- gitude. he possessed 50 years of good equinox and apogee positions. i. in other words. Similar calculations yielded the other two coordinates—heliocentric latitude. and calculate the perturbations of the radius vector and the lati- tude. First. consider first the calculation for the true longitude of the Sun for a given date. i. which expresses the amount of eccentricity of the elliptical orbit. then yielded the geocentric position of the Sun (or.190 T A B L E M A K I N G IN A S T R O N O M Y An example of the type of computation required in astronomy using these tables To illustrate die content and use of tables.e. using Delambre's tables. die secular variation (rate of change of precession per 'century). the mean longitude was corrected for the perturbations of die planets on the Sun. Now these values depended on past observations. and for perturbations on the Earth due to the planets (though the latter perturbations made a correction of only 2. the remaining two heliocentric position coordinates of the Sun. The result of these corrections was the mean time. to obtain the mean heliocentric longitude. one could look up the correction to the mean longitude for the true rime on the given day. But then further corrections due to changes in the perigee position had to be made to this value of the longitude. a correction to the mean longitude. 1805. In sum. and when Delambre drew up his tables of die Sun. in the case of the solar tables. Added to all this was the necessity to correct the obliquity. Finally. which made use of fifty entries from the tables.8 seconds out of 57 100 seconds in Delambtes tables).The data needed for these calculations appeared in 35 tables. and the logarithm of the radius vector (reduced to the ecliptic). 1805. The starring point for such a calculation was the mean longitude on January 1st of the year of interest.e. say 16h 7' 19" November 13th. the heliocentric position of the Earth). Next the change in longitude from January be to Ob 0' 0" November 13th was extracted from the proper tables and added. at the beginning of the nineteenth century. With these changes. another table yielded the equation of the centre. the true time had to be altered by the equation of time (the slowness or fastness of the Sun). From the value for the advancement of the perigee to November 13th. To this one added the indirect effect of the planets and the satellites. he generated a table of the mean longitude of the Sun for January 1st of each year. A simple conversion. a computer could generate an ephemeris containing heliocentric and geocentric position coordinates of the planet for as small an interval of time between positions as desired.

made between 1750 and 1762 at Greenwich.16 . and the necessity for an observer to correct observational positions for all these problems immedi- ately after they were taken.15 As a result. and the analysis of Bradley's observations appeared shortly after in 1818. nutation. At the suggestion of Wilhelm Olbers (1758—1840). Bessel set a new standard in the treatment of observations in order to improve the accuracy of their use in equations of motion. He also called attention to the treatment of instrumental errors. the positions calculated from these new tables showed substantial discrepancies after only a few years. outlined above. In all. at Greenwich. Bessel pointed out the necessity of continual observation of stars until enough information existed to predict their motion. Bessel reduced 3222 previously unreduced stellar observa- tions of Bradley. in place of occa- sional measurements that resulted in no sure comparisons. T A B L E M A K I N G IN A S T R O N O M Y 191 Problems with observed positions of planets and stars Turning now to the increase of the quality of observations. and aberration effects produced by the motion of the Earth. Bessel presented a new theory of precession in 1815. like their predecessors. These discrepancies could have arisen in two ways: either the observations were inexact or the theoretical developments were incomplete. the work in the eighteenth century of James Bradley and Nevil Maskelyne. Using these observations. a reinvestigation of the treatment of observations became necessary.14 For his analysis. These star reductions provided a new reference system for making observations of any celestial object. when Bradley was Astronomer Royal. he determined the latitude of Greenwich for 1755 from modern observation and used Bradley*s observations and those of Giuseppe Piazzi (1746—1826) to calculate proper motions. As the theoretical developments. Friedrich W. increased the astronomer's ability to calculate positions more accurately. the tables developed with Laplace's programme exhibited greater accuracy in the predicted posi- tions. along with selected others. a significant asset to the unified theory search. Bessel (1784-1846) attacked this problem in a number of ways from 1806 until his death. we note that some of the best observations available at the beginning of the nineteenth century for use in the equations of condition came from. Bessel recognized that he must first analyse the theories of applying corrections to stellar observations for precession. Further. However.

because they resulted in new values for the mass of Saturn and of its ring. He has been seen as a man of little imagination. Besides his improvements of the theories of precession.2' His discussion of the corrections needed in the elements of the orbits of the planets Mercury. but his contributions to astronomy were large because of his demand for accurate data and his long tenure. The demands he placed on himself and the staff at Greenwich resulted in large amounts of data. Indeed.192 T A B L E M A K I N G IN A S T R O N O M Y Several specific aspects of Bessel's investigations were of great signific- ance to nineteenth-century astronomy. Airy suggested that this inequality might be large enough to account for the errors in Burckhardt's tables of the Moon. He noted that each observer possessed a different rate of perception when looking through a telescope observing the morion of a celestial object. Airy's effect on positional astronomy was similar to Bessel's. refraction. Bessel's investigations into per- turbations were also significant. Airy's corrections were subsequently incorporated into calculation schemes for the British Nautical almanac. . Bessel recognized another brand of error: the personal equation. as a result. in that he and his computers reduced large amounts of data that were useful in detecting difficulties in planetary theory.17 and aberration.1* These methods advanced accuracy and increased the ease of application. his penchant for orderliness was the bane of his computers. and Mars led him to conclude that the mass of Mars was too high by a factor of 30 per cent. Airy was exceed- ingly meticulous in his observations and reductions. when coupled to the effect of the Earth. These he ascribed to an inequality of long period that was produced by Venus. Venus. (Burckhardt had made corrections to Delambre's tables in 1812.) Airy also discussed the coefficients of the lunar equation and found that there were several anomalies in the mean longitude of the Moon's position. but not for Mars.20 In the following year.' * Another important observer in the Besselian vein. Astronomer Royal from 1835 to 1881. he obtained corrections to the elements of the Sun's apparent orbit in a discussion of 1200 observations made at Greenwich between the years 1816 and 1826. based on his researches of Saturn's satel- lites. Airy's corrections agreed with Burckhardt's for Venus. the methods were adopted everywhere. Bessel improved the methods of reduction and published them in 1830 in his Tabulae Regiomontanae. and that this rate should be accounted for when reducing observations. was George Biddell Airy (1802-1892). As early as 1827 he compared Delambre's solar tables with 86 solar observations and found that several elements required alteration.

a refocusing took place in the nineteenth century. after Airy's suc- cession. he recalculated the mass of Jupiter and the elements of the Moon's orbit and he analysed data to further specify the position of the ecliptic. Educated as a mathematician. Besides attending to observations of position. This task. However. where he introduced Bessel's methods for the reduction of obser- vation. and partly due to Airy"s background. and this led to a series of memoirs by Hansen and by Leverrier. Even before Airy took command of Greenwich. the Admiralty awarded money to Peter Andreas Hansen (1795-1874). during part of which time he served as head of the Cambridge observatory. which allowed the German astronomer to complete and publish his'Tables of the Moon' (1858). by rapid publication of reduced observations. Leverrier (1811-1877) and John Coach Adams (1819-1892) both used Greenwich observations in their investigations of Uranus. Urbain J. hence better observations.22 This was due to the Admiralty's interest in the data for navigational purposes only. J. but gradually over the eigh- teenth century scientific supervision of the data decreased. theory again became uppermost in their minds in an effort to preserve the notion of stability of the solar system. he transformed Greenwich into a highly efficient institution. The Greenwich Observatory was the principal world producer of positional data from its founding. Tables of the planets proved less useful as time passed and a general dis- comfiture about them infected gravitational astronomers. constituted a permanent set of positions for the correction of planetary theory. he published a number of papers on the improve- ment of instrumentation at Greenwich. Airy spent several years as a professor at Cambridge. he proposed the reduction of the mass of observations lying dormant since 1750. partly due to the heightened interest in tables for astronomical use.23 The Observatory added to this set yearly. he did provide the basic data for others. at least half of the observations Simon Newcomb (1835—1909) used in his analyses of planetary motion were made at Greenwich. Airy. . In the late 1830s. For example. begun in 1833 and com- pleted in 1846.25 With the improvements in reduction. Airy also made an important indirect contribution to lunar astronomy before 1860: through his intercession. With this training and his exceptional abilities as an administrator. Later. also made a number of studies into the values of certain orbital elements of planetary theory. T A B L E M A K I N G IN A S T R O N O M Y 193 Airy became Astronomer Royal and Director of the Greenwich Observatory in 1835. like Bessel.24 While he did not noticeably advance perturbation theory. came a consequent increase in the discrepancies between observation and theory. In addition.

during this investigation. He is also considered one of the founders of modern meteorology. which other nations used as a model for their activities. to treat the problem. 7. Leverrier published several articles on celestial mechanics in the 1830s. having reorganized the meteorological service in France. he often was at odds with the staff. Hansen developed a new method for ease in handling the higher-order terms that emerged from the series . Paris. An authoritarian head of the Paris Observatory. (Courtesy of the Academic des Sciences. taking on the task of recalculating all the tables for use in the French ephemeris.) The contributions of Hansen and Leverrier to planetary theory The early works of Hansen and Leverrier grew out of a renewed concern for the older problems of the irregularities in the motions of Jupiter and Saturn. Nevertheless. A graduate of the Ecole Polytechnique.6 Urbairt lean Joseph Leverrier.194 T A B L E M A K I N G IN A S T R O N O M Y Fig. Instead of employing Lagrange's method of arbitrary constants. The problem Hansen faced here was similar to that of Euler and Lagrange before him. he began a study of the coefficients of the perturbing function expressing the perturbations of Jupiter and Saturn on each other. and. In the late 1820s. and. Hansen took up the problem of Jupiter and Saturn. he invented a new method for the analysis of perturbations. espe- cially of orders higher than the first. he made one of the nineteenth century's most significant contributions in the articulation of perturbation theory. Not long after he turned to study of the planets. he achieved a great reputation among the world's scientists for his work. which marked him as an important astronomer.

T A B L E M A K I N G IN A S T R O N O M Y 195 expansion of the perturbing function R. and Neptune. he became director of the Paris Observatory. in calculations beyond a hundred years. the treatment of perturbations became more difficult. In fact. Leverrier was led to consider the difficulties of the motion of Mercury. especially the more difficult ones of Mercury. unified theory by the end of his life. Previously. it served as a prelude to a second memoir in which Leverrier expanded the coefficients of the equations to the third power. Ephemerides in the nineteenth century Before completing the history of the development of a unified set of pre- cepts. the mass of analyses left by Hansen and Leverrier set many of the problems Newcomb would later deal with at the American Nautical Almanac Office in his successful attempt to develop a set of precepts with common constants for all the solar system bodies.29 and it led the way for Newcomb later. These other activities prevented him from completing a. and the many controversies surrounding it and the plot- ting of its orbit. Though he did not provide a unified synthe- sis. In the late 1830s. but Leverrier found that after a short time they indeed became noticeable. in 1853. it is instructive to examine the similarities and differences among the . precepts that included collections of terms to the eighth power. his was path-breaking work. He attended to short-period comets.28 In a sense. the contribution of these higher order terms had been considered negligible. Leverrier. Uranus. Leverrier was repeatedly sidetracked onto other astronomical projects. Nevertheless.27 He reanalysed the earlier investigations of Lagrange on this problem and found that. This work led to a new set of precepts for Mercury's motion. he was involved in the dis- covery of Neptune. As higher powers were included in the expansions. he employed the latest observations.26This method became an import- ant element in the work of Newcomb later in the century. in computing new precepts for planetary orbits. Through his investigations on the secular inequalities. which due to their accuracy. While his analysis was similar to its predecessors (though he ignored the new developments of Hansen) in that his expansions were carried out only to the first power of the eccentricity and inclination. took up the same question of the stability of the solar system. in his first astronomical memoir. Lagrange's formulae drifted. increased the amount of calculation required by Leverrier and his staff in the production of new precepts and positions.

). plus correction factors. and tables for the semi-diameter and the heliocentric rectangular coordinates of the Sun. but as the radii increased and the arc trav- eled per day decreased. Basically. the others tabulated the constants for reduction (precession. however. though once again the French offered less information. and tables for nutation effects on these stars.e. nutation. Only the Americans offered an ephemeris of Neptune. These values were given for Paris mean noon (the beginning of the Julian day) and Greenwich mean and apparent noon (in separate tables). moon culminations (the crossing of the moon through certain meridians). the orientation of Saturn's rings and the apparent discs of Venus and Mars. a planet.30 The French almanac was meant to serve the nautical community. which is used in calculating longitude at sea). than the French. Moon. the lunar distances (distance of the moon from the Sun. in the respective almanacs. the RA and 8 of the planet.The Americans. while the British and United States ephemerides attempted to fulfill the needs of both the nautical and astro- nomical communities. and United States) resembled one another. or a fixed star. the semi-diameter. the intervals rose to 20 days. Bessel's formulae of reduction using the notation of the catalogue of stars of the British Association for the Advancement of Science (1845). While the French gave the apparent positions for a list of 100 principal stars. and Jupiter's satellites. the horizontal parallax. The ephemerides for the planets were essentially identical. and the time of meridian passage. the almanacs gave the time of transit of a planet.31 Information on the Moon's motions included the RA and 8 for Greenwich mean time. All the almanacs contained information on the phenomena for that year: eclipses of the Sun. The British and Americans gave more information for the fixed stars. French. the equation of time. and only the British prepared tables of the minor planets. Each contained the apparent right ascension (RA) and declination (5). information on two stars of Ursae Minoris. those used as the reference positions. British. For their respective meridians. first published in 1855. The inner planets' heliocentric positions were recorded at 5-day intervals. divided their ephemeris.196 TABLE M A K I N G IN A S T R O N O M Y computed tables of planetary positions in the nineteenth-century ephemerides. the three principal ephemerides (i. They printed a list of the latitudes and longitudes of the principal observatories and tables to convert . into two parts: one in Greenwich mean rime for the use of mariners and one section in Washington mean time for the benefit of astronomers in the United States. etc.

Certainly the newness of the office in the United States encouraged the staff to choose the latest modifications in theories for their calculations. For the inner planets. the following remarks are restricted to the British and American almanacs. The Americans.33 In 1855. Encke's (1791—1865) analysis of the transits of Venus of 1761 and 1769 to obtain the equatorial hor- izontal parallax (or angular diameter) of the Sun at the Earth's mean distance. For the horizontal parallax and semi-diameter of the moon the office staff corrected these tables slightly. The almanacs also included the necessary data for observing occultations (or the disappear- ance from. The bases for the lunar tables in these two ephemerides provide an inter- esting anomaly. . Therefore. ignor- ing the publications of foreign astronomers. generated manuscript tables incorporating the latest observations and reductions of constants. I hasten to add that this does not characterize these French works as defect- ive. merely not as good as they could have been. They followed Airy's theory. They employed the tables pub- lished by the Bureau of Longitudes. and Miles Longstreths corrections of the equations for the Moon's longitude. Plana's (1781—1864) lunar theory modified with Hansen's inequalities and Airy's corrections of the lunar elements. To this. Judged from another perspective. Benjamin Peirce (1809-1880) and Joseph Winlock (1826-1875) added Hansen's analyses of the secular changes of the Moon's mean motion and of the motion of the position where the Moon is furthest from Earth (lunar perigee). while the Americans adopted Francesco Carlini's (1783—1862) later tables. the British still employed for the ephemeris of the Sun tables devel- oped in Milan in 1833. which was based on Giovanni A. we should point out also that the improvement in a number of cases was marginal. on the other hand. TABLE MAKING IN A S T R O N O M Y 197 from mean solar time to sidereal time and vice versa. The posi- tions generated from this theory compared more favorably with observation than did the tables of Burckhardt. view) of stars as the moon passes in front of them. The French used almost exclusively the products of French astronomers. prepared by members of the Board. But in noting that they adopted the latest tables.32 and the catalogue of the British Association for the Advancement of Science. though again the French information was less complete. The British still used Burckhardt's 1812 tables of the Moon. A. that of the information on which the ephemerides were based. The mean positions of the fixed stars for both almanacs came from Airy's Greenwich 12-year catalogue. which contained Bessel's improvements. the French also came in a distant second. Both used Johann F.

Poulkova. Meanwhile. Few astronomers were willing to invest such monumental . Such a programme required much technical and organizational skill to make it a success. None of these men lived long enough to complete their planned work. and Peirce planned to remedy the situation by providing new tables. espe- cially of Uranus and Neptune. comparisons revealed steadily increasing defects in the tables in use. particularly Peirce and Sears Walkers (1805— 1853) theory and tables of Neptune. Leverrier. While Leverrier followed closely the Laplacian programme. with corrections from Bessel. Leverrier began. Delaunay. and Peirce's corrections for the perturbations of Neptune. Only the elliptical solution for the orbit of Uranus came from Bouvard. so as to improve significantly on the earlier works. and for the outer planets they used Bouvard's of the same years. Simon Newcomb and the completion of the planetary programme If an astronomer desired to obtain one set of elements useftil in all the orbits. As the observatories at Greenwich. For Mercury they used Leverrier's theory of 1848. Paris. The astronomical community praised the changes made by the American Almanac Office because of the more accurate predictions that resulted.198 T A B L E M A K I N G IN A S T R O N O M Y the British employed von Lindenau's tables dating from the 1810s. and to this the staff added Leverrier's corrections for the perturbations of Jupiter and Saturn. Airy. Hansen deviated from the Laplace methods in many ways. The ephemerides of Venus and Mars were updates of von Lindenau's tables. to the bulk already reduced. a new set of tables for the motions of each of the planets. In every case. though Leverrier came close. but they made efforts to correct them with the latest data. Haasen. Similarly. In the late 1850s. The Americans had recourse to many of the same tables. This greatly increased the number of computations required before the generation of each ephemeris began. he had to substi- tute and cross-substitute the calculated elements until one set satisfied as many past observations as possible. they developed manuscript tables out of the earlier published works. and Washington added more planetary observations. the tables of Jupiter and Saturn came from Bouvard. In the next few years only Hansen succeeded in providing new tables for the Moon. the British and French incorporated a few of these modifi- cations into their almanacs.

He had before him the works of Laplace (1798—1825). mathematics. As we have seen. Newcomb had not only the ability to manage such a task. he received the equivalent of a state funeral. Newcomb reached the height of his powers in astronomy in the decade when both Hansen and Leverrier died. Newcomb examined the three principal methods of developing the perturbative function. and of Hansen on the minor planets (1857). When he died in Washington. wherein method (2) was modified such that the motion of one of the planets is described in terms of the sweep of the radius vector using the mean angular velocity of the planet in an ellipse. of Leverrier on the perturbative function (1855). 7. and (3) the Hansen method. Leaders of various foreign governments repeatedly engaged his services in the development of scientific activities in their countries. education. TABLE MAKING IN A S T R O N O M Y 199 efforts Into a task whose end result might prove to be little better than those of earlier tables. who published works in astronomy. leaving an incomplete legacy for a unified set of precepts.) . also in Laplace's Mecanique. and science fiction. in 1909. of Hansen on Jupiter and Saturn (1831). even more than Leverrier. after which the final positions were referred to a fixed plane. (2) the older mechanical method.7 Simon Newcomb was perhaps the most internationally known scientist at the end of the nineteenth century. (Courtesy of the Library of Congress Manuscripts Division. he had to decide which method was best for computing the general perturbations. Rg. he had the ambition to contribute to astronomy in as significant a manner as possible. He was a prolific author. of Peirce on Neptune (1852). DC. When Newcomb began his studies of Uranus' motion in 1858. there were: (1) the analytic method of Lagrange. religion. psychology. economics. presented by Laplace and used by Leverrier.

In his early attempts on Uranus. and when numerical values calculated from these expressions were added to the elliptic values of the elements a new planetary position could be predicted. and the results integrated to obtain variations in the coordinates in terms of this mean motion. he decided to rederive the Neptune theory using the ten further years available. was a straight application of the Lagrange method of variation of arbitrary constants. (2) The integration of the equations of motion. Once again. The integration of the equations with these constants.200 T A B L E M A K I N G IN A S T R O N O M Y There was also a threefold choice of method to integrate the equations of motion. (3) There was Hansen's method. . in 1864. At the time Newcomb published his Neptune tables. from which new positions of the planet could be obtained. he expressed the equations of motion in terms of the longitude. which were based on insufficient observational data—Neptune having been observed for only eight years at the time the tables were cast. So. Newcomb ran into trouble when he tried to use the major theories available for Neptune. his method was almost identical to that of Leverrier. they represented the observed positions excellently. The American Nautical almanac and the British Nautical almanac adopted this computational method. (1) There was the method of the variation of arbitrary constants we described above. Newcomb followed Leverrier's tech- nique. almost exactly. He expanded the perturbative function in terms of the distances of the planets along their orbits (using as reference the node of intersection of the orbits of the two planets). Then. the logarithms of the radius vectors of the planets. and latitude of the planet under consideration. Alas. They were replaced by a new set of Leverrier's tables for Neptune published in 1876. radius vector. the tables themselves did not fare so well. which met newer observations better. which remained in use for the next 30 years. where the perturbations were expressed in terms of a disturbed mean radius motion in the instantaneous ellipse. yielded expressions for the variation in the coordinates directly. and the mutual inclination of the two orbits. both items (1) of the above choices. The integration resulted in an expression for the variation of the elliptic orbital elements due to the perturbations. including perturbative forces expressed in terms of the coordinates. which Newcomb performed before 1870.

Hill (1838—1914) was assigned the work of Jupiter and Saturn. secular perturbation studies. For the last two planets. Publications of the Astronomical Society of the Pacific 49 (1937). but. to improve its usefulness. (This photograph originally appeared in F. could spend some of their time on the new planetary programme.8 George William Hill was a very accomplished mathematician. Hill made significant contributions to lunar theory. TABLE MAKING IN ASTRONOMY 201 None of this satisfied Newcomb's ambition to complete the Laplacian programme. Newcomb divided the rest of the work into two parts: the inner planets and Uranus and Neptune. and to free some time of the computers for other work. 7. comet orbits. He coordinated this work with Newcomb's intentions for the entire programme. The new struc- ture of the Ephetneris required less calculation. Schlesirtger. Astronomical Society of the Pacific. New York. He worked for the Nautical Almanac Office. p. and the calculus of finite differences. His collected papers fill six substantial volumes. reproduced with permission of the Editors. Newcomb first proposed a revision of the American ephetneris. subjective geometry. so the office personnel. Copyright 1937.) . George W. he planned only on updating his earlier work. This was a massive task and it took Hill ten years to complete. along with some new help.5.34 Fig. he was free to generate the tables in his own way. Besides his work on Jupiter and Saturn mentioned in the text. essentially. but did his calculations on the farm where he was raised in West Nyack. In a new programme begun in 1878. He submitted to the astronomers of the United States fifteen suggested changes.

a unified set of precepts and constants for the planetary system—the Laplacian dream.3* . which he employed as the provisional elements in the develop- ment of his own theory. in short. Second. he discussed the computation of the gen- eral perturbations of the planets. partially reconstructed them. and recomputed the positions of the four inner planets and the Sun. and he did the same for Piazzis observations at Palermo for 1791 to 1813. His final results embraced a complete revision of the planetary orbits and the masses of the planets. he carried over Arthur Auwer's (1838-1915) reductions of Bradley*s observations of 1750 to 1762.202 T A B L E M A K I N G IN A S T R O N O M Y Newcomb's programme for the analysis of the motions of the inner planets can be summarized as follows. and BessePs reduction of the observations at Konigsberg from 1814 to 1845. To save time. discovered in 1854. To determine this value. One important element in sorting out the perturbations of the planets on each other was the mass of Jupiter. To these Newcomb added his own reductions of later observations made at over a dozen observatories. as well as the principal constants of reduction. The total number of observations ultimately used for the four inner planets was 62 030! Third. Newcomb decided to calculate places for the planets prior to 1864 from these tables also. He began with Leverrier's formulae. and nutation. and changing the latter from Greenwich mean noon to Greenwich apparent noon and then to apparent noon at the observatories whose observations were to be used. fourteen pairs of planets had to be considered. He re-reduced Maskelyne's Greenwich observations from 1765 to 1811. so Newcomb was able to use observations of it from 1854 to 1889. Newcomb calculated a set of elements. precession. For the four inner planets. This asteroid made a near approach to Jupiter in 1885. aberration. From these manuscript tables. he undertook the re-reduction of the older observations. since Leverrier's tables were in use in the ephemerides of Europe from 1864 on. Newcomb used the interaction between Jupiter and the asteroid Polyhymnia. and Airy's reduction of John Pond's (1767-1836) observations of 1812 to 1830. and a discussion of the later ones to reduce them to a uniform system. because of the effects of Jupiter and Saturn on the inner planets. such as those for changing the Sun's positions in longitude and latitude to right ascension and declination. and reached perihelion in 1888. First. such as solar parallax. Several supplementary tables were also generated. as well as their own interactions. as well as Leverrier's reduction of the Paris observations from 1800 to 1875. using modern data.

Anton Pannekoek. But this did not lead to common use of either Leverrier's or Newcomb's tables in the generation of ephemerides. (Norton. Today's almanacs are more accurate and easier to use. which offers a better context . Writing 100 years later. The changes resulting from the use of these new tools swept away the tables developed in the nineteenth century. In spite of a somewhat greater accuracy obtainable with Newcomb's tables. now embedded in the computer programmes. tables used in the construction of the British Nautical almanac were based on Leverrier's. Johnson Reprint Corp. the constants. London. (Allen & Unwin. Beginning in 1960. TABLE MAKING IN A S T R O N O M Y 203 Newcomb proposed that his new fundamental star catalogue and the orbital constants and masses of the planets developed in the American office be adopted internationally. 1966) presents a highly detailed account of the concerns and accomplishments of the theoretical astronomy shortly after the publication of the work of Laplace and Lagrange. The Norton history of astron- omy and cosmology. and these were made and adopted at a conference in 1961. reprint edition. Robert Grant in his History of phys- ical astronomy. and electronic computers allowed for computational methods that are more accurate as more terms can be used in calculating positions. again until 1984. New York. Evidence from satellite observations and computer calculations revealed necessary adjustments in the constants. having been endorsed at an international conference in Paris in 1896. and the masses developed in the American office were adopted universally by 1903. though not the spirit of the Laplacian programme. New York. only the American Nautical almanac employed them.. A history of astronomy. These astronomical constants and tables with occasional modifications remained in general use until the advent of electronic com- puters and Earth satellites in the 1950s. 1995). (London.37 Further reading Treatments of the history of astronomy in the past 150 years have reflected the emphases in astronomy at the time. 1961) focused more on the developments in astrophysics that happened after Grant's time. Leverrier's tables served as the basis for the Connaissance ties temps until 1984. satel- lites contributed to improved observational data.36 the tables of the Sun. While there was some bickering in the astro- nomical community at the end of the 1890s. 1852. A more up-to-date work is that of John North. thus mak- ing the endeavours of sailors and mariners safer and more effective.

(Paris. 6. called the aberration. 7. Cotter's A history of nautical astronomy. de Laplace. The direction of a celestial object seen from the Earth is not the same as the direction seen by an hypothetical observer on the Sun. Those interested in the details of astronomical techniques can consult. vol. Notes 1. The centre of coordinates is the centre of the Earth and the parallax must be accounted for by an observer on the surface. 'Sur les solutions particulieres des equations different celles et sur les inequalities seculaires des planetes. P. Dover. S. See also Note 8. especially those by C. 2. Morando on Laplace. A. A short history of astronomy. L. 1902). presenth. vol. For the history of celestial mechanics. B.. 1772. Cambridge. New York. 1867—92). pp. Turin.par divers s(tivans (often cited as Savants etrangers). 1968). 1749. Eider. 4. (reprint edition. Dirk Brouwer and Gerald M. vol. besides the Mecanique of Laplace. 3. New York.1776. especially the plane of the ecliptic or of the celestial sphere. vol. called the stellar parallax. (Cambridge University Press. de Laplace.772.1766—69. For a charming and telling presentation of events and personalities in astronomy of the nineteenth century see Simon Newcomb. . 343—77. pp. The line of nodes is that line joining the two points at which the orbit inter- sects a given plane. 4. and Morando on developments in celestial mechanics in the nineteenth century. geocentric parallax. (Paris). general history of astronomy. (Hollis & Carter. S. Methods of celestial mechanics (Academic Press. London. Reminiscences of an astronomer. another form of parallax. et sur les inegalites seculaires des planets qui en dependent. Recheirhes sur la question des inegalites du mouvement tie Saturne et de Jupiter. 1961). 1961).204 T A B L E M A K I N G IN A S T R O N O M Y for the developments in the old and new astronomy. P.'Sur la methode des variations.' Mem. 2. (This essay was begun in 1772. *Sur le principe de la gravitation universelle. L. Technical details of navigational astronomy can be found with ample diagrams in Charles H.' Misc. Lagrange. 1995) are a necessity. New York. 5. Clemence. The finite velocity of light gives rise to an apparent motion of a star or plan- etary object as the Earth turns in its orbit. results from the finite size of die Earth.Wilson on Lagrange. For a good brief his- tory consult Arthur T. from which the discussion of Hansen's work in this essay was taken. Centres de Lagrange. 37-63.' Memoires de mathematique et de physique. Paris. Hoskin (general editor).) . Berry. 2. (Riverside. Soc. published in 1775 in the volume for year 1. J. the elegant essays in M. read in two parts (1773 and 1774).

'The solar tables of Lacaille and the lunar tables of Mayer. See also. see E. A popular history of astronomy during the nineteenth century. Traite. 1830. 199-276. 97—102. p. Refraction of light from a star or planet is a change of direction in the light's path as it moves from an area of one density of matter to another. T. ES. E W. 2.Taton and C. E. 12. Cambridge. J. Maskelyne for use in calculating lunar positions. Courcier. 11. pp.The eighteenth and nineteenth centuries (ed. Maskelyne published an English corrected set of the tables to honour Mayer. 'Sur 1'alteration des moyens mouvements des planetes. Lagrange. A. 20. 169-292. 255—71. Bohn. 8. required by Mr. Paris. Airy. 15. 13. 1798-1825. J. J. Mayer sent manuscript tables to N. *Sur les equations seculaires du mouvement des noeuds et des inclinations des orbites des planets. pp. 117 (1827). 1899. Tables astmnomiques. 635—709. Planetary astronomy from the Renaissance to the rise of astro- physics: Part B. 16.Wilson).* Memoires Academic des Sciences (Paris). 1806. 383—430. 1770. Duprat. L. Wilson. . Chez J. Lagrange. 19. 'Theorie des variations seculaires des elemens des planetes. Bessel. by Tobias Mayer. 5. results from the grav- itational attractions of the Sun and Moon on the asymmetrical Earth—it is pear shaped. 65-70. The line of apsides is the major axis of an elliptical orbit.. 1781.1782. At this time. London. 344. 5 vols. Clerke. B. Precession. 'Novae tabulae motuum soils et lunae. History of physical astronomy. W. Forbes and C.' Dictionary of scientific biography. Lebon.. 'Friedrich Wilhelm Bessel. L. 9. Histoire abregee de I'asttonomie. 4th edn. vol. vol. pp.' in The general history of astronomy. Volume 2. L. de mecanique celeste. Premiere Partie. vol. See New and correct tables of the moon and Sun. R. M. 4. 'Remarks on a correction of the solar tables. 14. The effects of the Sun and Moon on the Earth produce another smaller wobble of the polar axis called nutation—an oscillation of the pole about the position it would have if only precession were present. pp. 2 (1753). 1852.' Mem. 1774. Henry G. Regiomonti Prussorum. Oeuvres de Lagrange.' Notweaux Memoires de I'Academie Royale des Sciences et Belles-Lettres (Berlin). B. pp. 1776. After Mayer's death.' Philosophical Transactions of the Royal Society of London. 10. 1995. 17. 125-344. London. 6. Tabulae regiomontanae reductionum observationutn astmnomkamm ab anno 1750 usque ad annum 1850 computatae.' Commentarii Sorietatis Royale Scientiarum Gottingensis.1972. Lagrange. London. R. Mayer. For a fall treatment of Mayer's work. Oeuvres de Lagrange. Academic des Sciences de Berlin. 18. pp. Oeuvres de Lagrange.. such as it does when it enters the Earth's atmosphere. vol. South s observations. T A B L E M A K I N G IN A S T R O N O M Y 205 7. 1902. Paris.de Laplace. Cambridge University Press. G.. Grant. akin to a wobbling of the Earth's polar axis. Bureau des Longitudes. M. Gauthier-Villars. Fricke. Paris.

at the Royal Observatory. Venus. Hansen. 'On the corrections in the elements of Delambre's solar tables required by the observations made at the Royal Observatory. . 33. aus dem Venusdttnhgange von 1761 hergeleitet. Airy. Eggin. 23—34. Cambridge. pour les sept planetes principales. Airy.'George Biddell Airy. Washington. beginning in 1855. 1896. London. A brief summary of the work and some results was published separately. 1. Autobiography of Sir G.J. 1824. 'Tables de la Lime constroits d'apres le principe newtonien de la gravitation universelle. DC. 209—19. 1-66. 34. Paris. 24. Akadeinie derWissenschaften zu Berlin. Newcomb retired as head of the Almanac Office in 1894. 28. Catalogue of 2156 stars. Airy. 22. B. G. see E A.J. Paris. Cambridge University Press.J. American ephemeris and nautical almanacfor the year [date] (beginning in 1855). Der Venusdurchgang von 1769. U. Hansen. Clerke.'Sur les variations seculaires des elements des orbites. from 1836 to 1847. Connaissance des temps ou des nwuvement$ celeste. Leverrier. 1849. Unterstuhungen iiber die gegenseitigett Stonmgen desjupiters und Saturns. Leverrier's many results can be found in the volumes of the Annales de 1'Observatoire de Paris. 1845. B. Mercure. US Government Printing Office. 32. The American office was established in 1849 and published its first ephemeris for the year 1855. Die Entfernung der Sonne von der Er. 35. 30. Gotha. Theorie du mouvement de Mercure. 29. U.* Dictionary of scientific biography. Leverrier. For details of the method. F. 1840. 23. Saturne et Uranus. 31. 84-7.Washington.'JoMma/ des mathematique pures et appliques 3 (1858). begin- ning with volume 1 in 1882. ah Fortsetzung davon. Mars. Gotha. The elements of the four inner planets and the fundamental constants of astronomy (Supplement to the American ephemeris and nautical almanac for 1897). vol. 1822. Greenwich. pp. US Government Printing Office. P. 79. see W. but continued on to complete his work on the unified planetary theory with additional appro- priations from the US Congress. (beginning in 1767) London. Astronomical Ephemeris and nautical almanac for [year]. S. Popular history (note 19). la Terre.206 T A B L E M A K I N G IN A S T R O N O M Y 21. a I'usage des astronomies et des navigateua [year) (beginning in 1679). J. Berlin. Paris.J. O. 26. Greenwich. Jupiter. Encke. 27. 1895.' Philosophical Transactions of the Royal Society of London 118 (1828). J. Newcomb and his colleagues published the results of this work in a new pub- lication of the (American) Nautical Almanac Office called the Astronomical papers prepared for the use of the American ephemeris and nautical almanac. 1970.' Additions a la connaissance des temps pom Van 1843. 1831. p. formed from the observations made during twelve years. G. For Airy's bibliography. A. Newcomb. Airy. 25. DC.de.

Washington. 1983. L. J. Doggett). 37. T A B L E M A K I N G IN A S T R O N O M Y 207 36. L. S. Dick and L. DC. US Naval Observatory. Details of the controversy are discussed in A. I am grateful to Dr P. E.* in Sky with ocean joined (ed. Frana for help with preparation of the figures. .'Simon Newcomb's role in the astronomical revolution of the early nineteen hundreds. Norberg.

Ab—fld ALPHABETICAL LIST OF OCCUPATIONAL TERMS SHOWING THE CODE NUMBER OF THE HEADING TO WHICH EACH IS TO BE REFERRED. 879 .

8 The General Register Office and the tabulation of data. (Ministry of Labour. rela- tionship to head of family. birthplace. marital condition. HMSO. London. 1837-1939 EDWARD HIGGS In the 1861 census of England and Wales householders were asked to give eight pieces of information on each member of their household—name.) . age. 8. profession. sex. A dictionary of occupational terms.1 Page from an occupational dictionary used by the General Register Office in 1921. Rg.

but the cen- sus authorities. and what it does not mean. The construction of forms and tables involves decisions about what is important and what is not. But in this case it was not what the census authorities wanted. and modern historians. an. Similarly. Data truncation in a world of scarcity Jasper Messenger's 'incorrect' census schedule entry points to the central paradox of all census-taking. or non- kin terms implying a pecuniary status: 'servant'. as well as in the nineteenth century. and that person is someone in a posi- tion of authority. It also involves a decision about what information means. if one looks in the nineteenth-century Census reports. described his relationship to the household head AS 'friendly*. agricultural labourer living in a shed at the bottom of a farmers garden in the village of South Cerney in Gloucestershire. house- holders had printed instructions on their household schedules. 'friendly' does not appear on any of the axes of tables relating to relationship to head. and 'boarder*. To the bureaucratic mind we can be male or female but nothing in between. Certainly. reflecting the changing concerns of such people.210 THE G E N E R A L REGISTER OFFICE and certain information on medical disabilities. or having a little joke. and what should be collected and pre- sented. or by ethnicity. . Someone makes decisions about what is 'important'. Citizens are not asked what they find important. To help them. But why is the term 'wrong'? After all a 'friendly relationship* is a perfectly sensible English usage. or what to them reveals the 'state of the nation'. the rest is regarded as irrelevant to the task at hand. and who we love is irrelevant. would have almost certainly said that his entry had been filled in 'incorrectly*. demands only certain pieces of informa- tion about individuals. along with many other social processes—that they are as much about the truncation of information flows as their collection. However. This made it quite clear that 'relation- ship to head of family' was to be understood in terms of kinship. But these decisions can also change over time. A census schedule today. Jasper Messenger. 'lodger*. the two-way tables given in nineteenth-century Census reports select merely a few vari- ables to hold up for inspection: occupation cross-tabulated by sex and age but not by the social class of ones mother. 'apprentice'. and what can be ignored. and an example of how to fill in the form.1 Jasper may have been foolish.

1861. 259.2 Search Room of the General Register Office. The GRO was established in 1837 to supervise the secular system for reg- istering births. 8. In the case of the organization considered here. Aprii 20. THE GENERAL REGISTER OFFICE 211 However. p. (Illustrated Times. analysis and presentation needs to be cur- tailed to fit the resources available. information collection. the General Register Office (GRO). They may merely reflect the need to do certain tasks in as efficient a manner as possible given the scarcity of means. Such scarcity of resources applies as much to the public that has to read tables as to the institutions constructing tables. there is nothing necessarily sinister about such processes of truncation within survey methodologies. The GRO maintained a central 'database' of copies of the certificates of registration of Fig. and deaths set up locally in England and Wales under the provisions of the 1836 Registration and Marriage Acts.) . since the human ability to digest information is limited. the resources available for the processes of tabulation were indeed limited. 1861. marriages.

which helped underpin advances in both medical science and public health. and so on. and on the characteristics of individuals already noted above. and cross-tabulated in various ways. regional. was dis- seminated in published parliamentary papers. gathering information on. from which it produced weekly. and national levels. on housing.2 From 1840 the GRO was also responsible for the administration of the decennial censuses. The processes of collecting and analysing census and civil registration data involved the handling of vast amounts of information. 1838-79. (The Mansell Collection Ltd. household structure. the size and distribution of the population. 8. annual. A key element in this process was the analysis of cause of death data supplied on death certificates. The census produced import- ant statistical information in its own right but it was also the basis of much of the GRO's medical and demographic work—knowledge of the size of the population allowed the calculation of national. During the English . was at the forefront of the development of medical and sani- tary statistics from his appointment in 1839 to his retirement in 1879. William Farr. age and occupational figures allowed the calculation of mortal- ity statistics and life tables for age cohorts and particular trades.3 Dr William fan. quar- terly. local.212 THE GENERAL REGISTER OFFICE Fig. This material.) vital events issued by local registrars. aggregated at the district. The GRO's first superintendent of statistics. and decennial reports on medical and demographic trends. and cause-specific death rates. compiler of statistics and superintendent of statistics at the General Register Office.

have marked against their names in the private register kept as to their conduct and qualifications either 'indifferent'. However. In order to undertake this work the GRO had precious few hands. This was extremely fiddly work. and the established staff of the Office. and the figures were then read off on to district sheets. whilst registered vital events in the same year totalled just over 2000000. Figures were transferred from district to county sheets in a similar manner. For most of the nineteenth-century the GRO had fewer than 90 staff. the sheets for sub-districts had to be folded at the column to be totalled and then lined up so that they overlapped. for example. giving the raw numbers of people under particular occupational headings within particular age ranges. 10 pieces of information were gathered on 32 527 842 individuals. These headings were ruled across the sheet. Until 1911 the GRO's clerks had to add up and present results via the use of tabling sheets and the 'ticking' method. and two of those died. and the constant leaning over tables was fatiguing for . or 'bad' or 'very bad'. The ticks in the columns were then added up. temporary clerks were the dregs of the clerical profession over whose selec- tion the GRO had little control. and the results placed in another series of columns on another sheet.* These temporary clerks. The quality of the clerks can be judged from an official complaint to the Treasury in 1889 that in the previous census: There were altogether 98 clerks supplied to the Census Office. or more than a half. Of these 98. THE G E N E R A L R E G I S T E R O F F I C E 213 census of 1901. Moreover. by Treasury nom- ination. In the case of occupational abstraction by age. the tabling sheets were large pieces of paper with occupational headings down one side and age ranges across the top. only had simple. and of these never more than 25 percent were employed in the Statistical Department which undertook the tabulation of data. creating a matrix of boxes into which the census clerks were to place a tick for an occurrence in the enumerators' returns of a person of the relevant age and occupation. Of the remaining 94.4 The Office had made similar protests to the Treasury in 1861 and 1871. for example. no less than 49. Sheets were created in this manner for each registration sub-district. independently ot some few writers transferred from the GRO.* In order to create tables for registration districts. four were in such bad health that no work could be got from them. since in the nineteenth century the necessary Census Acts were usually passed only some nine months prior to Census Night.3 At the time of the census the GRO was able to employ temporary clerks to undertake the work of data processing. there was little time to organize and train such staff". manual technologies to undertake this vast task of data processing and presentation.

the Treasury expressed itself very doubtful about making an appointment at all. and there was plainly a danger that data presented in a complex manner would baffle and confuse them. on the not insub- stantial grounds. necessitating frequent stops for rest. Forms of data truncation in the GRO Given these constraints the GRO required simple and easily understood means of presenting data. could not afford to fail to get its message across. However. the first registrar general. approached the Treasury in 1838 for per- mission to employ someone for the 'difficult and important duty' of draw- ing up an abstract of the causes of death.7 In 1856 the GRO reported to the Treasury that an experienced clerk had just completed 'tabling the ages and diseases of the females of Lancashire 1854. But the adoption of this essentially two-dimensional form of representation entailed a necessary simplification of the complex natural and social phenomena which the Office sought to study. Thomas Lister. and in this the two-way table was its prime statistical device. These forms of . that the Treasury grudg- ingly gave way.063 "ticks"'. with a distinct scientific and political agenda. It was only because Lister complained that he had already prevailed upon the Presidents of the Royal Colleges of Physicians and Surgeons.214 THE G E N E R A L REGISTER OFFICE the (abler. processes as simple as possi- ble in the nineteenth century. A newly formed institution such as the GRO.9 Civil servants.8 This cumbersome process helps to explain why the GRO sought to keep the census and civil registration. and that this had taken him four days. and the Master of the Society of Apothecaries to circulate medical practitioners with an exhortation to provide accurate information on the death certificates for statistical purposes. and why its statistical output was restricted to tables with simple cross-tabulations. At its inception the GRO was also under pressure to produce quick and easily comprehensible results because the Treasury was initially doubtful about the usefulness of the Office having a statistical function at all. the medical profession and the general pub- lic were woefully innurnerate then as now. as being otherwise calculated to mislead in a matter of very great importance'. it would be better that it should be omitted. 'that unless the information proposed to be given can be afforded with the certainty of entire accuracy. comprising 310 abstract sheets containing an aggregate of 29.

that helped to form cells whose meanings appeared unambiguous. and the assumption of homogeneous populations. People could then be neatly placed once. but also by limiting the numbers and scope of the questions asked. the GRO claimed that it had no legal authority to collect such information. and no attempt was ever made by the Office to obtain information about morbidity (i. from the decennial enumeration. In the case of the census. the numbers of cases in various census tables would be the same. Many did so. Census respondents. thus avoiding difficult questions over comparability.e. which may have precluded the recording of much casual. On the death certificate. its parent department. to do so. and thereafter the GRO fought to prevent other questions being inserted into the schedule. dur- ing the processes of tabulation.12 . especially that of women and children. usually the first. and the lead in the subsequent insertion of new questions was taken by other government departments. the questions asked were almost identical from 1851 through to 1881. although the census clerks were instructed to only abstract the 'most important*. Profession or Occupation'. As a result. Taking these in turn. and once only. and seasonal work. for example. in forms such as 'Farmer. sickness rather than death—'mortality'). for example. in one of the boxes.10 Even in the twentieth century the number of questions asked by the British census was always significantly lower than that in the US census. THE G E N E R A L R E G I S T E R OFFICE 215 complexity reduction can be summarized in the following manner: the restriction of data collection to a narrowly defined set of questions. all that was asked for was cause of death. the reduction of multi-faceted variables to unitary entities. It had to be ordered by the Local Government Board. partly via the development of the forms and sched- ules used in the census and civil registration process. unitary entities.11 The use of simple two-way tables also encouraged the Victorian GRO to present variables as undifferentiated. When a Treasury Committee on the Census in 1890 supported calls from social scientists for the insertion of a question on employment status. The wording on the census schedule also spoke of people having a 'Rank. although they were encour- aged to give multiple occupations on their census schedules. could only have one occupation in the published tables. the information collected by the GRO was inten- tionally circumscribed. Maltster and Brewer*. the subsumption of complex responses under general headings in clas- sification schemes informed by ideological assumptions.

These all worked by placing reported causes of death and occupations under broader headings that were then the terms placed on the axes of tables. citizens were presented as living simple lives. was the cancer or the irritant the primary cause?14 Just as the census clerks had rules for choosing one occupation amongst many.. then those of the longest duration were to be given precedence. on their death certificates doctors defined 'primary* in several ways. Allied to these issues were the conceptual difficulties arising from the elaboration of various classification systems in terms of medical nosologies (classifications of diseases). and diabetes in IV.13 The later introduction on the certificate of the concept of a'secondary' or'contributory' cause of death caused even more confusion. scarlet fever in II. and socio- economic groupings. as by tar or oil. Chronic diseases were to take precedence over non-chronic. and dying uncomplicated deaths. for example. could be referred to some definite irri- tation. Where multiple causes of death came from the same group. influenza in III. for example. for instance.216 THE G E N E R A L REGISTER OFFICE Similarly. These difficulties led the GRO to begin to doubt the very objectivity of the concept of a primary cause of death. As it noted in the Statistical renew for 1927 it was very difficult to say. violence was to be preferred to general diseases as a cause of death. and local diseases to ill- defined causes. general diseases to local diseases. in any instance. According to the 1911 instructions to the GRO's clerks. rather than from multiple causes. since it then became difficult to tell whether any 'secondary* cause was regarded as a consequence of the primary. and if all else failed the first given was to be taken as the cause of death for the purposes of abstraction. came in Group I.15 As a result of these processes of truncation. but contributing appreciably to the death. occupational classification systems. However. or in terms of the most important with regard to the termination of life. or as of independent origin. enabling the dead to appear only once in each annual table. If a cancer. either chronologically. all the various jobs and occupational titles involved in the production of cotton goods would be subsumed under the broader headings such as . so the registration clerks were given rules for choosing which causes of deaths were to be given precedence. the mortality tables in the GRO's Annual reports assumed that everybody died of a 'primary* cause of death. how far the quest for the 'primary cause' should be pushed. In the occupational classification system. Terms for general diseases were then placed into four classes in order of preference for selection: smallpox.

8. 1927.4 Part of a GRO table showing deaths by age in 1871. 1871.«—e»a««n«<4 Fig. LOHDON. p. London. HMSO. (Annual report of the Registrar- General (Abstracts of 1871).) . 156. «(5 Causa of Death.—CAUSES of DEATH at aifeeat Periods of Utt in tie Yew J871—Mu.

whether peritonitis? or enteritis? 80. 69. and Causes of Death. Broken. heart . Sometimes . Abdominal inflammation (vague. 104b. Bronchial ulceration extending to the lungs 69. . Calculus (urinary) Stone. Bruised corn 144. Carbunculus Carbuncle. 30f. Abortus Abortion. Ciccutn (stricture of) 84. . Bursting of a fallopian tube Under what circumstances? 104. Canker Of what part? Cancnuu oris? 39d. Phlegmon? Bursa (inflammation of) 105. Bronchi (abscess of) (bad) Phthisis? Pneumonia? 71—73. Cancrum oris 39 d.81. Bronchi (inflammation of) Bronchitis. For the names which it is recommended should be used.) Poisoning? of stomach and intestines Ulceration? 137 85. 17. 132. Carbuncle 132. Bursting of a blood-vessel See 28. The queries in the second column will remind the informants and registrars of points which should be borne in uiiitd in assigning the causes of death. Cachexia Scrofulous? Cancerous? 33.144.218 THE GENERAL REGISTER OFFICE An Alphabetical List of Diseases. see the first column of the previous nosology* to which the figures always refer. Abdominal effusion Ascites. Abrasion of the mucous membrane (A bad term.35. 69. Bronchocele 33. Bronchitis 69.73.) Query. . . Cancer 35. nearly atl of which have been met with in the Registers— with References (by figures) to the Statistical Nosology The insertion of names in this list must not be considered as by any means sanctioning their use. Bronchitis and broken rib How was the rib broken? 69. Bronchial fever Influenza? Bronchitis? 13. 97. improperly used for rupture of the heart. Caecum (biliary) 89b. Mental distress. Of what kind? of what part? Cancer (chimney-sweepers') 35.

THE G E N E R A L R E G I S T E R O F F I C E 219

Carcinoma Cancer, 35.
Carditis 58,
Caries Of what bone? i l l .
Catacausis 40.
Catalepsy 54.
Catarrh 69.
Catarrah of the bladder 98b.
Catarrahal fever Influenza? 13.
Catarrahus epidemicus hifluerr/a, 13.
Catarrahus vesicae Catarrah of the bladder, 98b.
Cauliflower polypus of the womb 37c; 103.
Cellular dropsy 30.
General Register Office, 4th annual report of the registrar general of births, marriages
and deaths for 184O-41, HMSO, London, 1842, pp. 166-86.

'Cotton Goods Manufacture'.This allowed the size of the tables to be radi-
cally reduced, thus aiding the processes of tabulation and public
understanding. But the use of such systems inevitably introduced into the
tables certain ideological principles that have bedevilled subsequent analysis,
The GRO's system of socio-economic groupings in the early part of the
twentieth century highlights these issues. T. H. C. Stevenson, the GRO's
senior medical statistician from 1909 to 1931, constructed this classification
in order to study the class-specific fertility of married women, as revealed
by questions in the 1911 census on completed family size. He assumed that
the social status of a family could be determined by assigning it to one of
a series of socio-economic groups on the basis of the occupational desig-
nation of the male household head. The classification was based on the
concept of a social gradation of skill and intelligence which was assumed
to be reflected in occupational biographies, and to be constant both spa-
tially and over time. The taxonomy used was basically tripartite at it formal
inception, with 'the upper- and middle-class' in Class 1;'those occupations
of which it can be assumed that the majority of the men classified to them
at the census are skilled workmen' in Class III; and Class V containing
'occupations including mainly unskilled men'. Two groups of intermediate
occupations were then placed between these classes, making five in total.
Thus, the term 'physician* would be placed in Class I, 'vet' in Class II,
'cotton weaver* in Class III, and 'general labourer' in ClassV. Stevenson used
this classification to 'show' that marital fertility declined the 'higher* one

220 THE G E N E R A L REGISTER OFFICE

progressed up the social scale. He explained this in terms of the gradual dif-
fusion of knowledge on mechanical contraception from the middle classes
'downwards' in society. The GRO's system was subsequently taken up
widely within the social science research community as a means of stratify-
ing households into social classes.'6
This system of socio-economic groupings based on job titles, however,
masks all sorts of local and occupational anomalies in family size, and obscures
the dynamics of fertility decline. These cannot be simply read off from occu-
pational designations but need to be seen in terms of the negotiations which
went on between men and women within families over sexual and reproduct-
ive matters. It was not levels of intelligence which determined fertility but
concrete matters such as the availability of work for married women, the
relative cost of having children, and the gendered balance of power within
marriage. The assumption that the occupation of the male 'breadwinner',
rather than total family income, determined status, and the lumping together
of various occupational groups with very different social and economic experi-
ences, have all been criticised by historians.17
Similar issues could be raised with respect to the occupational classifica-
tion systems used by the GRO in the Victorian period. These were not based
on self-evident and concrete economic structures but arose from particular
intellectual models which appeared relevant at the time. The occupational
classification systems were based on the grouping of occupational titles
according to materials being worked up. These, in turn, were taken as affect-
ing the morbidity and mortality of those who undertook these occupations.
In addition, William Farr also appears to have followed classical precedents
in believing that working with particular materials affected the character of
workmen. Those who failed to indicate the material upon which they
worked in the census schedule, or those, such as clerks, who did not shape
or tend particular substances or animals, were fitted into the material-based
classification in residual categories. Thus "labourers' who neglected to indic-
ate their branch of employment were abstracted under the heading'General
Labourers'. This occupational category included over half a million men in
1871. Similarly, clerks, warehousemen, and many dealers, were placed in the
material categories relating to the establishments within which they worked.
Thus a clerk in an iron mill would be abstracted under the heading 'Iron
Manufacture Service1, rather than under'Commercial Clerk.* The retirement
of Farr, and his replacement as superintendent of statistics by William Ogle,
led to an important shift in the principles of occupational classification used
in the 1881 census, with clerks, and other tertiary workers, now being

THE G E N E R A L REGISTER OFFICE 221

abstracted under their own distinct headings (see Fig. 8.1 (a) at the beginning
of this chapter). This shift may lie behind the apparent tertiary economic
'revolution* in the late nineteenth century.18
Similarly, what the right to life meant in practice was constrained by how
the GRO and other public health bodies perceived death. As Christopher
Hamlin has recently argued, the early public health movement under Victorian
reformers such as Edwin Chadwick can be seen as an attempt to constrict
what health implied. Rather than conceiving of the individual manifestations
of disease in traditional terms as the impact of a myriad of external influences
on unique human constitutions, Chadwick followed Neil Arnott and Thomas
Southwood Smith in seeing disease as the result of the invasion of the body
by specific chemical pathogens. This allowed Chadwick to narrow the con-
cept of a right to health to the right to the removal of noxious human efflu-
ent from the cities, and to clean water supplies. Chadwick could thus outflank
medical and Chartist claims that health had to be seen in broader constitu-
tional terms such as access to food, rest, and tolerable working conditions.19
Farr agreed with Arnott and Southwood Smith that infectious diseases
were caused by chemical blood poisoning. The whole cast of the medical
nosologies drawn up by the early GRO, which placed each death in a single
category according to the 'primary' cause of death, also helped to narrow
how the concept of mortality could be understood. It was no longer pos-
sible to use terms such as 'Cold', or 'Damp clothes (putting on, or sleeping
in)', as a cause of death—terms such as pneumonia and bronchitis needed
to be substituted. In order to enforce this new concept of specific causes of
death, which was ideally suited to the creation of the distinct cells of the
two-way table, the GRO issued doctors with its nosologies, and sent death
certificates back to them if they did not conform to the new form of clas-
sification. Plainly, this revolution in medical classification has brought vast
benefits in terms of diagnosis and therapeutics but it has also detracted from
the ability to see patients as whole human beings placed in a social setting.
To exaggerate wildly, people do not have diseases because of the conditions
within which they find themselves, they have become the medium through
which causes of death manifest themselves.2"
Equally problematic was the assumption made in the Victorian period
that the populations in each of the cells of two-way tables were homogeneous.
A central aim of the GRO was to create tables showing the number of deaths
in distinct administrative districts. When combined with, data from the cen-
sus on the population of such units, this would enable the calculation of
death rates, usually expressed as so many deaths per 1000 live population,.

222 THE G E N E R A L REGISTER OFFICE

These figures were then used to identify places with 'excessive' mortality,
i.e. that in excess of a 'Healthy Districts Death Rate* based on a group of
agricultural districts with low mortality. This provided ammunition for
local sanitary and public health reformers. Under the 1848 Public Health
Act, the central General Board of Health could also force a community to
set up a local board of health if its annual death rate was above 23 per thou-
sand. Crude death rates of this sort were easy to calculate from tables of
mortality, and easy to understand. They were, however, deeply flawed, since
they took no notice of the unique age and sex distributions of individual
communities. This was especially serious in the case of resorts with high
proportions of elderly residents, and relatively high mortality, whose local
councils were highly critical of the GRO's tables.21

Transcending the limitations imposed
by data truncation

Much of the subsequent technical history of the GRO can be seen in terms
of attempts to overcome the limitations imposed by these forms of informa-
tional truncation through technological and methodological innovation.
The former helped to speed up analysis and computation, and so allowed
more, and more complex, tables to be created. Methodological innovations
were introduced, however, in order to make the tables work better, or to
transcend them altogether.
The general reluctance of the Victorian GRO to expand its data-processing
operations stands in marked contrast to its willingness to experiment with
new forms of computational technology. The GRO did not try to find new
ways to reduce the mass of individual survey returns to quantitative results
but it did attempt to find ways of manipulating the latter more easily when
the old manual systems of data processing had created them. As Williams has
described in Chapter 5, Georg Scheutz, a Swede with a background in
printing and publishing, was attracted to the idea of producing error-free
text via perfecting the printing unit of Charles Babbage's difference engine.
The machine developed by Scheutz and his son was brought to England in
the 1850s, won the approbation of Babbage and a committee of the Royal
Society, and went on to win a gold medal at the Paris Exhibition of 1855.
This apparatus was obtained by the GRO in 1858, at the considerable cost
of ^1 200, for the purpose of generating the actuarial tables for the Office's

§2 EUSHSM UKE TABU, Ho. 3,

Fig, 8.5 A page from the GRO's English Life Tables of 1864. (General Register Office,
English life tables, HMSO, London, 1864, p. 52.)

224 THE G E N E R A L REGISTER OFFICE

English life tables."The machine's capacity to replace both human 'computers'
and compositors, and thus produce reliable, error-free aids for the insurance
industry, was seen as its most important advantage. By replacing clerks with
machinery it was also hoped to halve the cost of producing tables. George
Graham,, the head of the GRO from 1842 to 1879, claimed in a letter to the
Treasury in August 1860 that the machine had produced 153 pages of tables
since February of that year at a cost of £77, whilst it would have cost £,153
if the work had been done by two clerks.22
William Farr expressed bis admiration for Scheutz's invention in his intro-
duction to the English life tables but it is plain that in action the machine was
very temperamental. Michael Lindgren, drawing on Scheutz's own correspon-
dence, claims that the use of the machine was, on the whole, a failure, since only
a quarter of the English hfe tables were calculated and printed with the machine.
But even this rate of success was in fact beyond the capacity of Scheutz's inven-
tion since the Stationery Office informed the Treasury that only 28 pages of
the completed work were entirely composed by the machine; 216 pages were
partly so composed; and the remaining 520 pages were entirely set up by hand.
This, and its limited application, almost certainly explains why the use of the
difference engine was not taken up widely within government or commerce.
Although the GRO continued to utilize the Scheutz machine for at least the
next twenty years, the amount of tabulation it could perform was plainly lim-
ited by its specialist function and its unreliability,23
The real innovation in the GRO's computational arrangements came
with the introduction of commercial mechanical calculators. The first com-
mercially manufactured calculating machine, the 'arithmometer', had been
designed by Thomas de Colmar in 1820. However, according to Andrew
Warwick, although de Colmar's machine was exhibited and awarded a
medal at the Great Exhibition of 1851, the large-scale mechanization of
computation in British commerce appears to have only begun in the late
1860s. The first public debate in Britain on the utility of the arithmometer
to the human 'computer' took place in the pages of the Assurance magazine
in the early 1870s, The GRO's purchase of an arithmometer in 1870 for
j£20 was thus an. extremely early example of the application of this techno-
logy in Britain. William Farr informed the Treasury that the machine had dou-
bled the number of calculations a clerk could make in a set time, whilst
enhancing their accuracy. By 1872 the GRO was obtaining Treasury per-
mission to buy a second arithmometer, and further machines were purchased
in 1873, 1877, 1881, 1893, and 1903. The GRO's use of such technology
was far in advance of some other comparable government institutions.

THE G E N E R A L REGISTER OFFICE 225

According to Mary Croarken, for example, the Nautical Almanac Office at
Greenwich was performing most of its computational work for astronom-
ical tables by manual methods into the 1920s. It only began to introduce
calculating machines in 1912,24
In the 1890s the GRO also began to employ various forms of slide rule,
which the Office found both more useful and cheaper than arithmometers.
Thus, in a letter of August 1895 the registrar general informed the Treasury
that all calculations in the Statistical Department were made twice—once
with an arithmometer and once with a slide rule. He continued that, 'it is
only by the adoption of every possible mechanical appliance, for the saving
of clerical labour, that the present staff of that Department can keep pace with
the constantly increasing demand for calculated rates and percentages*.25
However, throughout the Victorian period the GRO stuck doggedly to the
manual'ticking' system of data processing. This technology was coming under
increased strain in the early years of the twentieth century, as new questions
were added to the census. The GRO was also anxious to ask the new ques-
tions about marital fertility already noted. In order to analyse the fertility data
in the 1911 census, and that gathered by the other new census enquiries, the
Office introduced Hollerith tabulators. These had been invented in 1890 for
the US census of that year, and were being introduced into state statistical
offices across Europe. This was a consequence of the increasing size and com-
plexity of national census enumerations across the Western world as a whole
in a period of widening state intervention into society.26
Machine tabulation broke data analysis down into two stages. First,
information on individuals was punched on a card as a series of holes, and
secondly, the information on the cards was read electronically. In essence,
pads with spring-loaded pins were brought down on individual cards, and
if the pins passed through a hole they completed a circuit through which
electricity passed to move the dial of a counter on one position. The dials
could be wired up via relay keys to undertake complex calculations. What
this did was to separate data capture from data analysis, since the cards could
be analysed in differing ways, and as many times as required. At a stroke the
bottlenecks in the GRO's manual system of data processing were removed,
opening up whole new possibilities for statistical manipulation. Although
introduced for census purposes, the new machines also allowed the GRO
to generate new medical tables, such as those showing the relationship
between primary and secondary causes of death, and the places where
people died.27 The invention of the 'database' thus preceded that of the
electronic computer by more than half a century.

8. Scientific American. 30 August 1890. (Cover.) .Fig.6 The use of Hollerith machines in the US census of 1890.

an increase of 20 per cent in the number of maternal deaths registered as from causes associated with childbirth other than puerperal ('childbed') fever in the period after 1927. published in 1885. for example. this was at the cost of portraying death as a state rather than as a process. William Ogle. John Tatham.28 In essence the GRO was admitting that its published cause of death tables were merely one set out of a whole series of possible tables. his successors as superintendents of statistics in the GRO had to grapple with the deficiencies in his chosen tabular methods. This process of transcending the legacy of Farr was carried still further in the 1930s by the introduction of the modern forms of rnultivariate analysis associated with Francis Galton. In his Letter in this volume. the standardization of populations. However. Karl Pearson. In 1927 the GRO changed the instructions for the order of the statement on cause of death in the death certificate. Ogle's successor as superintendent of statistics. and it was introduced into the Annual report of the registrar general for 1901. There was. This then provided a more definite starting point for an analysis of the train of related causes. the GRO's superintendent of statistics from 1880 to 1893. rather than the best of all possible representations. for any others of which it was the consequence. The numbers of deaths by place and cause were also increasingly adjusted by basing their calculation on a standard population with a defined age and sex structure. THE G E N E R A L REGISTER OFFICE 227 Perhaps less well known are the methodological innovations pioneered by the GRO in the late Victorian period and in the early decades of the twentieth century to overcome some of the limitations of the two-way table. began to use such standardization in the Decennial sup- plement covering 1871—80. The GRO believed that this innovation led to significant changes in reporting. These can be seen in terms of the reconceptualization of death. This made comparisons of crude death-rates with those of earlier periods misleading. Tables thus came to present worked rather than raw data. in order.29 Although William Farr has rightly been lauded as the father of modern medical statistics. The concept of the 'primary' cause of death had allowed Farr and his successors to create two-way tables that claimed to reveal the face of death at any particular point in time. noted that the more general introduction of death rates standardized by age had been made necessary because of changes in the structure of the population con- sequent upon the decline in the birth-rate in the late nineteenth century. and the introduction of rnultivariate ana- lysis. and George Udny . by calling for the 'immediate' rather than the primary cause first. and then.

Stocks continued working at the GRO until 1950. geographical latitude.32 Paradoxically. so allowing . Pearson. This was consequent upon the appointment of Dr Percy Stocks as 'medical statistical officer* at the GRO in 1933. and socio-economic class. infusing the older statistical tradition of Farr with a new rigour and analytical power. Stocks started out his career as a house physician at Manchester Royal Infirmary. The Review published in 1938 contained a discussion of the need to introduce measures of statistical sig- nificance in those cases where the number of cases of a particular cause of death was under 20. housing density. had finally arrived in the GRO. In the Registrar general's statistical review for 1934. although in the 1940s he was increasingly drawn into the work of the World Health Organisation (WHO). and plus or minus 3 per cent at WOO. These allowed the Office to control the amount of information it had to process. The probability statistics of Galton.3' Survey methodologies and information truncation The story of table-making in the GRO reveals a general pattern-—the reduction in informational complexity and the subsequent reintroduction of such complexity via improved technology and methodological advances. and their successors. from 1921 to 1933.228 THE G E N E R A L R E G I S T E R OFFICE Yule. published in 1936. and was assistant school medical officer in Bristol in the period 1918 to 1921. Stocks showed that the standard error decreased as a percentage of the number of deaths by cause as the number of cases increased—plus or minus 32 per cent with 10 cases. before going on to become Reader in Medical Statistics at the Gallon Eugenics Laboratory under Pearson. the triumph of the GRO in the Victorian period lay not just in collecting and displaying 'data' but also in finding ways of truncating the collection and presentation of informa- tion via forms and techniques of tabular presentation. The same Review also saw the introduction of local areal comparability factors (ACFs) to allow standardization between localities. at least at first sight. Stocks introduced the use of correlation coef- ficients and regression analysis to show the relationship between mortality. In the years immediately pre- ceding his appointment to the GRO he was a regular contributor to Pearson's Annals of eugenics but his work was mainly epidemiological rather than overtly eugenic.30 The advent of Stocks was associated with an increasing statistical sophis- tication in the GRO's work.

to the presentation of a partial view of the world. Edward Higgs. and Simon Szreter. 435—64. inevitably. and thus necessitate further innovation. London. Johns Hopkins University Press. 115-34. we advance as much by creating interesting problems as by finding solutions to them. pp. T H E G E N E R A L R E G I S T E R OFFICE 229 it to create a coherent picture of the certain aspects of the external. 'The GRO and the Public Health Movement in Britain. London. London. Eyler. HMSO. and the methods of data abstraction and tabulation used by the GRO in their analysis. class and gender in Britain 1860-1940. Thus. provides similar context for the GRO's development of its socio-economic . 1996. Victorian soda! medicine. HMSO. 4. Simon Szreter. 11. The standard text for the career and work of William Farr is John M. 1979. vol. But this is not a unique history. Cambridge. A dearer sense of the census: the Victorian census and historical research. which others wanted supplemented or expanded. world. Further reading A general history of the GRO can be found in Muriel Nissel. Edward Higgs. pp. gives a general introduction to the nineteenth-century cen- suses. 1800— 1854. see Christopher Hamlin. since all human development can be seen in terms of solutions to problems that create their own drawbacks. Continuity and Change. Social History of Medicine. 1987. Public health and social justice in the age of Chadunck: Britain. They also allowed the GRO to present information in a simple form that could be comprehended by a busy public. For the specific circumstances of its creation in the early nineteenth century see. 1996. vol. 1996. and intellectual context of the GRO innovations in disease classifications. These techniques allowed great advances in survey methodologies but also created considerable problems for subsequent analysis. 1998.1991. administrative. Fertility. Cambridge University Press. 'A cuckoo in the nest? The origins of civil registration and state medical statistics in England and Wales'. More details on Farr and several other of the figures mentioned here can be found in the Dictionary of National Biography and the standard obit- uaries of the period. For the early nineteenth-century political. Cambridge. The truncation of information flows involved in the GRO's survey methodologies led. 1837-1914'. This led to further technical and metho- dological change within the Office's techniques of statistical production. People count. Cambridge University Press.

RG9/1783. 1996. 78-81. and 15 September 1890. esp. pp. Campbell-Kelly. Stanford University Press. Cambridge. p. PRO.. People count. PRO. 10. 'Social monitors. This is a very difficult text but one that merits the effort. Stanford. Higgs. Szreter. California. 8. Eyler. Notes 1. 6. Bulmer (ed). London. Johns Hopkins University Press. pp. class and gender in Britain 1860-1940. 127. London. esp. technological innovation and the production of medical statistics*. 'The statistical Big Bang of 1911: ideology.. MH 19/195. 155. 7. 1995. in M. RG 29/3. A dearer sense of the census: the Victorian census and historical research. 155-6. E. 1985. For William Farr see. 39—51. The data processing implica- tions of this innovation can be found in Edward Higgs.35. M. Public Record Office (hereafter PRO). Cambridge. Nissel. Cambridge University Press. M. M. C. pp. The role that the truncation of communications plays in social systems is discussed in Niklas Luhmann. Local Government Board and predecessors: Correspondence with Government Offices. enclosure. Social History of Medicine. 5. 9. 2. esp. Continuity and Change 11 (1996) 115-34. . T 1/6028B/12646. 409—26. 1987. 'A cuckoo in the nest?'. 3. Fertility. 11. The GRO instituted simple tests to weed out the worst of the Treasury nominees but the overall standard of the original population was inevitably low. Higgs. 27. 1979. Essays on the history of British sociological research. For the history of the GRO see. Cambridge University Press. A dearer sense of the census.230 THE GENERAL REGISTER OFFICE classification in the early twentieth century. HMSO. Higgs. pp. Census Returns. J. 22-36. 1996. E. Higgs. pp. 1861. 119—20. Hakim.1996. vol. 'A cuckoo in the nest? The origins of civil registration and state medical statistics in England and Wales'. GRO Letter Books. population censuses as social surveys'. London. Social systems. Treasury Board Papers. f. S. letters of 30 August. 4. p. HMSO. 'Information technology and organizational change in the British census. 41. 123. Victorian soda! medicine. 1801-191 \\ Information Systems Research 17 (1996). 8 September. Martin Campbell-Kelly has argued that the GRO could control the selection of the census clerks via the system of Civil Service examinations but this is to contuse the recruitment for established posts within the Gifice with the tem- porary employment of casuals in the Census Office. 9. PRO.

Princeton University Press.600. Making sense of the census. General. 1. Warwick. 1. 18. p. MIT Press. English life table. Cambridge University Press. 249-50. Treasury Board Papers. Norton-Wise. 50 (1976). forthcoming. in M. 1993. cxlii. An example of the tables produced by the Scheutz machine can be found in PRO. p. E. GRO Letter Books. Lindgren. pp. GRO Letter Books. . RG 29/2. 1. 328-9. The qualitative dimension of quantitative demography. 274-86. letter of 31 May 1872. Early scien- tific computing in Britain.. pp. Sholkamy (eds). in S. London. Dhartnalingatn and H. Office of Population Censuses and Surveys and predecessors. PRO. C. p. Glory and failure: the difference engines ofjohann Mutter. 233. GRO Letter Books.11-2. Register Office. 52-66. General Register Office. PRO. cxli. 421. 20. 24. A. program. RG 29/5. Registrar general's statistical review for 1. pp. by W. p. 1990. M. Szreter. M. 28. 1929. Szreter. Princeton. 16.W Cortada. Parliamentary Papers (hereafter PP) 1893-4. 179. 67-282. Fertility. 129-30. 1864. RG 29/7. 1990. 6-7. p. GRO Letter Books. XI. General Register Office. NCR. with an intr. tables of lifetimes. 145. E. IBM. 1800-1854. pp. Higgs.10-20. 1801-1901. Princeton University Press. p. 13. Eyler.927.T l/7299b/11516. RG 29/3. 538-46. London. Before the computer. xvii.The Office obtained £12 from the Treasury in 1877 for its repair: PRO. p. Oxford University Press. 300. Higgs. 19. A. PRO. Rules of Selection from Jointly Stated Causes. Cambridge. PRO. 80-9. Fart. RG 29/2. First and second report of the select committee on death certification. PRO. 161—6. PRO. 'The linguistic construction of social and medical categories in the work of the English General Register Office'. RG 29/6.Treasury Board Papers. London. Burroughs. 80.'Mortality statistics and Victorian health policy. 23. A dearer sense of the census. and Remington Rand and the industry they created. 101. Oxford. Charles Babbage and Georg and Edvard Scheutz. GRO Letter Books. pp. 22. Hamlin. HMSO. GRO Letter Books. 15. London. pp. G. Croarken. HMSO.1. M. J.* Bulletin of the History of Medicine. 22. T 1/6257B/12792. p. 21. RG 29/6.149.. p. PRO. HMSO. Princeton. pp. The values of precision. Public health and social justice in the age of Chadwick: Britain. THE G E N E R A L R E G I S T E R OFFICE 231 12. annuities and premi- ums. 17.864. 1989. 'The laboratory of theory or what's exact about the exact sci- ences?'. p. Higgs. RG 29/6.. pp. p. Ibid. class and gender in Britain 1860-1940. Cortada also sees the emergence of a European market for calculating machines as being a feature of the 1870s:J. RG 26/88 Classification of Cause of Death. General Register Office. Clarendon Press. 1995. pp. Statistical Branch: Population and Medical Statistics: Correspondence and Papers. 1998. 1865—1956. 250. The manuscript returns for England and Wales. 335-55. and crit- icism. RG 29/1. Oxford. 443-602.4. English life tables. pp. pp.

Black. 27. HMSO. Higgs. London.'The statistical Big Bang of 1911. 1. . with fixed sex and age distribution. 4—8. 9-15. London. 145. births and deaths 1911-1920 HMSO. 1954. Luhmann. London. Registrar general's statistical review for 1934. 83-5. Registrar general's statistical review for 1936.232 THE GENERAL REGISTER OFFICE 25. 30. 'Proposal for the establishment and international use of a standard population. 409-26. 1929. 93. pp. London. Stanford. pp. London. p. 28. 1989. 1938. Clarendon Press. 64th annual report of the registrar general for 1901. 149. pp. 760-1. General Register Office. 31. England and Wales 1921. Social History of Medicine. Rome. HMSO.934.103. 1971-1980. p. pp. Stanford University Press. statistics of marriages. A business and technical history. ideology. PRO. HMSO. General Register Office. HMSO. General Register Office. Registrar general's statistical review for 1. including anaemia and measles. 29. London. p. Imperial calendar 1934. Oxford. HMSO. London. GRO Letter Books. London. General Register Office. pp. Ogle. 1933-1936. HMSO. E. General Register Office. 169. birth and death rates'. technological innovation and the production of medical statistics*. p. General Register Office. Registrar general's statistical review for 1936. 1903. pp. Medical Research Committee and Medical Research Council: Files. General Register Office. M. xxxiii-xxxvii. p.A. Who was who. 122. 275. Bulletin de {'Institute International de Statistique. 1912. London. xvi. & C. 1981. 150-5. 8—13. pp. PRO. General Register Office. See also W. extract of the Committee meeting of 26 October 1934.951 Text. The registrar general's decennial supplement. see N. Part III. 1995. ICL. FD 1/7110 MRC Statistics Committee: Report. in the calculation and comparison of marriage. For a general theory of systems development along these lines. RG 29/3. 1933. 1936. Social systems. 1938. London. 26. HMSO. Estimates of population. Registrar general's statistical review for 1927. Seventy-third annual report of the registrar general for 1910. 190. vii—viii. RG 29/7 pp. Campbell-KeBy. 9 (1996). 32. HMSO.Vl (1892).

This page intentionally left blank .

Table XIV—Pertamna Function 56 .

) . their typography and high standard of production were directly attributable to L }.1 British Association for the Advancement of Science Mathematical Tables Volume l. (By kind permission of the British Association for the Advancement of Science. the Committee's most influential Secretary. Although the tables were the result of a collaborative effort by members of the Mathematical Tables Committee. 9 Table making by committee: British table makers 1871-1965 MARY CROARKEN The British Association for the Advancement of Science was set up in 1831 to promote the public understanding of science. By the 1870s it had come to play a central part Fig. The first separate volume of mathematical tables produced by the British Association Mathematical Tables Committee was published in 1931. Comrie. It set a standard of accuracy and ease of use which later volumes of tables strove to achieve. an activity in which it is still has an important role today. 9.

was just beginning to establish his reputation as a mathematician and table maker.236 TABLE M A K I N 6 BY C O M M I T T E E in scientific life in England. Glaisher.Thomson and Stokes. a committee was set up for the purpose of reporting'on Mathematical Tables. In 1870 the British Association allocated . which it might be desirable to compute or reprint'. Mathematical tables were the main computing tool for physicists. the more specialist mathematical tables were often published in mathematical and scientific journals.2 The Committee con- sisted of Arthur Cayley. As a result it was becoming increasingly difficult to trace tables and to avoid the unnecessary labour of recalculating tables already in existence. engineers. and two leading British mathematical physicists. The lack of a method of identifying tables was a primary motivation for the creation of the Mathematical Tables Committee3 but so too were the mathematical interests of some of the original committee members. Henry Smith. Indeed Glaisher took the view that 'one of the most valuable uses of numerical tables is that they . Sir William Thomson.£72850 today. a student of Cayley's. The annual meetings. the British Association also supported new areas of scientific invest- igation through the creation of temporary research committees. chairman. held in different cities around Britain and occasionally overseas. The concerns which lay behind the creation of the Mathematical Tables Committee were based on the increasing amount of computation being required in scientific research especially in the physical sciences.£1472 in research grants worth approximately . The research committees were intended to undertake clearly defined tasks over a period of a few years and report their findings at the annual meetings. to engage in debate and to socialize. etc. It contained two of Britain's leading mathematicians. and James Whitbread Lee Glaisher. and trigonometric tables were commercially produced and easy to obtain. The meetings were an opportunity to listen to scientific papers. Cayley and Smith. namely James Whitbread Lee Glaisher and Arthur Cayley. and mathe- maticians during the nineteenth century and had become increasingly numerous and diverse. Some commit- tees were supported by small grants to defray the costs of whatever investiga- tion was being made. Sir George Stokes. logarithm. were regularly attended by eminent scientists and mathematicians as well as by their less well known colleagues. transactions of societies.1 At the 1871 meeting of the British Association held in Edinburgh. While standard collections of 4 to 6 figure multi- plication. secretary. pamphlets. As part of its mission to promote the understanding of science.

was the driving force behind the early work of the Mathematical Tables Committee. William Thomson. and Henry Smith (clockwise from top left) constituted the first Mathematical Tables Committee. Stokes and Thomson. George Stokes. W.2 The British Association Mathematical Tables Committee 1871. 9. TABLE MAKING BY COMMITTEE 237 Fig. Cayley and Smith. The Committee represented two of Britain's leading mathematicians. Arthur Cayley (Chairman). L. His first task was to draw up an extensive catalogue of general mathematical tables based on a classification . and enable the extension of the former to bear fruit practically in aiding the advance of the latter'.) connect mathematics and physics. J. with guidance from Cayley. and two of Britain's leading mathematical physicists.4 Establishing the Mathematical Tables Committee. Glaisher (Secretary). 1873-88 Glaisher. (Thomson photograph with permission from the Institution of Electrical Engineers Archives and the others with permission from the London Mathematical Society Archives. 61aisher was a young man at the start of his long career as a Cambridge mathematician.

early 1876 producing 30 manuscript volumes at a cost of . Glaisher had hoped that the Mathematical Tables Committee would be able to publish any tables they produced in book form as he perceived one of the main difficulties with using mathe- matical tables was their dispersal throughout the periodical literature. Glaisher soon began to make and publish tables both on his own and on behalf of the Mathematical Tables Committee. The catalogue was published in the British Association annual report for 1873. Glaisher and his father supervised a team of eight computers who worked on the tables from October 1872 to. Glaisher published die tables independently in 1883.5 It was a landmark in mathematical table bibliography and remains an authoritative source on tables available in the nineteenth century. In 1872 Glaisher designed the computation and supervised the calculation of the tables by a freelance computer paid from a British Association grant to the Committee. At the 1871 British Association meeting Glaisher read a paper on the calculation of e and expressed his intention of calculating tables ofe* and e~x for which he hoped to gain financial support from the British Association. Another table of the same function was independently calculated and Glaisher himself compared and differenced the two tables to detect and elim- inate as many errors as possible.8 The project was a huge undertaking requiring the calculation and differencing of 64 800 tabular entries. Cayley was also in the process of preparing a book on the subject.238 T A B L E M A K I N 6 BY C O M M I T T E E system devised by Cayley. It was supplemented two years later by a catalogue of number theory tables compiled by Cayley. approximately. However by 1879 it had become obvious that such a project was too ambi- tious and the Legendrian Function tables were published in the 1879 British Association report. Glaisher had been introduced early to table making by his father. considered important because of their application to mechanics and to interpolation. The project was never taken up by the Mathematical Tables Committee but may have been a contributing factor in arguing for the creation of such a committee. head of the magnetic and meteorological department of the Royal Greenwich Observatory and an experienced and competent table maker.7 As early as \ 872 the Mathematical Tables Committee took the decision to systematically tabulate elliptic functions. In addition to his bibliographic work.6 The first tables which Glaisher made on behalf of the Mathematical Tables Committee were Legendrian Functions. Elliptic functions had many applications in mathematical physics and this project had the backing of the whole committee. James Glaisher.

L. Rayleigh had had first hand experience of publish- ing Bessel function tables in his Theory of sound12 and knew that such tables . the subject which dominated the work of the Mathematical Tables Committee during the late nineteenth century was Bessel function tables. The tables were published in three volumes in 1879. Alfred Lodge. The next project undertaken by Glaisher on behalf of the Mathematical Tables Committee was the calculation of factor tables—the first of many number theory tables to be published by the Committee. James Glaisher Senior joined the Mathematical Tables Committee in 1877 specif- ically to oversee the work using two hired computers paid for by a grant of £250 from the British Association. W. TABLE M A K I N G BY C O M M I T T E E 239 around £450 to the British Association. and sixth millions. Proceeding slowly: Bessel functions and number theory tables 1888-1928 In 1888. and retained Cayley.. The British Association spent another £250 on printing proof copies of the tables. and Thomson. The Mathematical Tables Committee decided to fill the gap by publishing tables of the fourth. Lord Rayleigh. and Bartholomew Price. 1880. and his later work on the history of factor tables. was never held to account. Alfred Greenhill. The committee's activ- ities were strongly influenced by Rayleigh and Lodge who served as Chairman and Secretary respectively. Factor tables are the most universally useful of all number theory tables and Glaisher's 1873 catalogue.10The publication of the factor tables effectively ended J. fifth. the Mathematical Tables Committee was reappointed as a 'Committee appointed for the pur- pose of considering the possibility of calculating Tables of certain Mathematical Functions'. five years after the factor tables were published. Glaisher.William Hicks. as far as can be estab- lished. never took the final step of publishing the tables and.11 The committee had five new members. Glaisher. however.9 had identified factor tables covering numbers up to three million and numbers from six million to nine million. Bessel functions were used in many branches of mathematical physics to describe physical phenomena and were perceived as an entirely appropriate subject for the Mathematical Tables Committee to address. and 1883. Glaisher's involvement with the Mathematical Tables Committee although he continued nominally to be a member until 1901.

rather he did much of the computation himself in his spare time with the help of two volunteer computers (one of whom was his sister Eleanor). factorization. Unfortunately the new machine proved mechanically unreliable which seriously slowed down the work and Lodge had to borrow a machine once again. Lodge did not ask for large sums from the British Association. and received. Lodge asked for. In 1889 he used an Edmondson calculating machine to check the printed tables by continuous addition of the first differences. The tables were pub- lished in 1900 with financial help from the Royal Society. to farther speed up the computations. The tables were published piecemeal in the British Association annual reports with the intention of gathering them all together to publish in book form at a later date—a goal which was not achieved until 1937. and the solution of binomial congruences to base 2) which he had begun to compute. In 1895 Lt. Lodge was the first to apply a calculating machine to the Committee's work. Colonel Allan Cunningham turned the Committee's atten- tion temporarily away from. Throughout that time he proved to be an extremely diligent computer contributing many tables to the British Association Reports. Lodge remained a member of the Mathematical Tables Committee until his death in 1937. Bessel functions by suggesting that the Mathematical Tables Committee publish a table of residues of powers of 2 (useful for testing divisibility of numbers.240 TABLE M A K I N 6 BY C O M M I T T E E were needed by mathematical physicists in many fields. The Committee agreed to support the project and Cunningham took over from Lodge as Secretary in 1896 specifically to further the publication of his work. j£15 with which to employ a professional computer. carried out the work on a much smaller scale than either Glaisher's elliptic function or factor table projects. In the early 1890s. By 1899 the work was complete and Cunningham asked the British Association to fund the printing costs but they granted only half of the required amount. Over the next few years the Mathematical Tables Committee began to calculate and publish Bessel function tables. For the next four years Cunningham dominated the Committee. which paid for the computation of a second table to compare with the first and for the comparison of the two copies. Initially the machine was borrowed from a colleague but later the Mathematical Tables Committee acquired a machine for its own use. . Lodge who undertook most of the computing during this period.13 There is no doubt that Cunningham used the Mathematical Tables Committee to fur- ther his own reputation and to finance the publication of his tables.

and Ronald Fisher. Thompson 1948 J.J. in the main. his student Louis Filon as Secretary and. J. Lodge 1888-1896 W-Tfaomson 1897-1901 A. the only link to the old Committee.R. Arthur Doodson.Nicholson 1920-1931 J.R.M. Only 250 were printed and almost half of these were donated to libraries and mathematical institutions.W. For the next twenty years the Committee gradually expanded to contain many of Britain's best mathematical table makers including John Airey.Darwin 1951-1956 D. promptly left the Committee once his tables were published. The Committee published a steady stream of tables in the British Association annual reports but there was no strong direction or leadership of the Committee. TABLE MAKING BY COMMITTEE 241 Chairmen and Secretaries of the Mathematical Tables Committee Chaimum Dates Secretaries Dates A. H. All work was carried out on a voluntary basis by Committee members and sometimes unsystematically and. P.J.W. Wilkes 1958-1969 * For a short while there were two Mathematical Tables Committees but they were soon amalgamated. Lodge. C. G. In 1913 it reverted to the tide Mathematical Tables Committee to reflect its members' growing interest in tables other than Bessel functions.Airey 1920-1929 E.j. G. Initially the Committee consisted of just three members. Cayley 1871-1889* J. Mordell 1948-1950 Royal Society 1948- Administrators C. C. Demand for the tables turned out to be low.W. Comrie 1929-1937 J.L. Micaiah Hill as chairman.Nicholson 1910-1920 J. Following a six year period of inactivity the British Association Mathematical Tables Committee was again reconstituted in 1906-—this time with a specific remit to tabulate Bessel functions. Filon 1906-1910 J. Hill 1906-1919 L.Glaisher 1871-1889* Lord Rayleigh 1888-189? A. Neville 1931-1947 L. J.V. Cunningham 1896-1901 M.Hartree 1957 M.Wishart 1937-1946 A. Miller 1946-1948 L. . Cunningham. N.

C. the most influential person on the Committee was the Secretary. Building a reputation. 394 with permission from the Mathematical Association. Alfred Lodge. New York: Yeshiva University. Leslie Comrie. Of the eight people to hold the position following Glaisher's initial work. There appears to have been no cohesive policy to find out which tables were needed by the wider community in order to target Mathematical Table Committee activity.) . and Jeff Miller (clockwise from top left) were the most significant. Allan Cunningham. One of the Fig. John Airey. 14. (The photograph of Lodge is taken from Mathematical Gazette vol. p. 1928-39 At the 1928 meeting of the British Association in Glasgow six new mem- bers were appointed to the Mathematical Tables Committee. The other photographs are taken from K.242 TABLE MAKIN6 BY COMMITTEE Committee members made tables which were of specific interest to them. Scripta Mathematica Studies No. 1948. Until 1948. Archibald. when the Committee was taken over by the Royal Society. 93 Selected Secretaries of the Mathematical Tables Committee 1888-1948. 3. Mathematical Table Makers.

was establishing himself as Britain's leading table maker and expert on machine computation.16 The British Association had financially supported the publication of Volume 1 of the Mathematical Tables Series but could not commit itself to financing expensive publications on a regular basis. TABLE M A K I N G BY C O M M I T T E E 243 six. however. the International Astronomical Union asked the Mathematical Tables Committee to consider publishing a table of Emden functions. that most of the tables needed recalculating to get them all into a consistent format and to allow for interpolation between the tabulated values. The first volume of the British Association Mathematical Tables Series was published in 1931. He soon found.15 Although edited by John Henderson. vision. The second volume of British Association Mathematical Tables was a joint venture between the Mathematical Tables Committee and the International Astronomical Union. then Deputy Superintendent at the Nautical Almanac Office. Leslie John Comrie.17 Comrie had great respect for both men's table making ability and trained them in his table making philosophy where attention to detail and a focus . the book was a true cooperative venture to which many Committee members directly contributed. and set the standard for a series of tables which became a byword for accuracy and ease of use. and succeeding vol- umes. The Bessel function tables also needed systematic recornputation so these were left for a later volume. In 1930. Comrie. His first task was to publish a collection of tables based on those published in the British Association annual reports.14 On a professional basis he supervised the pro- duction of the Nautical almanac and was engaged in a number of commer- cial table making projects. and mechanical computing methods. Comrie undertook to compute the tables of second order differen- tial equations and used the opportunity to bring Donald Sadler and Jeffrey Miller onto the Committee. the British Association Mathematical Tables Committee built up a world wide reputation.Through it. Comrie was fortunate in being able to gain sponsorship from other sources. professionalism. Comrie very soon pruned the Committee of older members and organ- ized other Committee members to be responsible for different aspects of the work. Sadler was Comrie's deputy at the Nautical Almanac Office where Comrie was now Superintendent and Miller was the man Connie would have chosen for the post if ill health had not prevented it. a specific form of differential equation used in the investigation of stellar structure. By 1929 Comrie had been elected Secretary of the Mathematical Tables Committee and brought with him vitality. had a profound effect on the work of the Committee.

Use of calculating machines for scientific computation was spreading and Comrie was one of the leading advocates of machine computation. and a National Accounting machine with the money. cycles of reduced ideals in quadratics fields. were donated to the Committee in the late 1930s. and Factor table giving the complete decomposition of all numbers less than 100. The British Association National was installed at the Nautical Almanac Office where Comrie and his staff soon developed very efficient differencing techniques21 which he applied to British Association work. pre- pared by Glaisher at the end of the nineteenth century. which had already declined to publish the Glaisher elliptic function tables when they came to light. After extensive consultation with some of Britain's leading mathematicians19 the Mathematical Tables Committee used the money to publish volumes entitled Minimum decompositions into Jifth powers. The Committee. Comrie had also been buying calculating machines for the Nautical Almanac Office but had been unable to persuade the Admiralty to finance the purchase of a National Accounting Machine. . Comrie purchased a Brunsviga-Dupla. Manuscripts of number-divisor tables and a table of powers.4. The Cunningham bequest was also used to support the publication of Volumes 8 and 9 of the British Association series which owed their exist- ence to another past member of the Mathematical Tables Committee. or other computers. While Miller was the better mathematician. Glaisher. Cunningham died in 1928 and left ^3272'8 to the Committee to be used to produce new number theory tables. as needed for the work of the Committee.244 TABLE M A K I N 6 BY C O M M I T T E E on usability dominated. The other machines were held variously at the Nautical Almanac Office and at the Rothamsted Agricultural Experimental Station under the supervision of Ronald Fisher and were used by Committee members. After leaving the Mathematical Tables Committee Cunningham had continued to work on number theory tables and he also left money to the London Mathematical Society for a similar purpose. a Nova Brunsviga. and 5 of the British Association Mathematical Tables Series were made financially possible by an earlier member of the Committee.000. Allan Cunningham. Volumes 3. Sadler had the ability to organize computing projects and to deal with typogra- phy issues.w Another important benefit of the Cunningham legacy was the financial freedom to purchase calculating machines for use by the Mathematical Tables Committee. Both Miller and Sadler became dominant members of the Committee for the next 30 years and witnessed great changes in the com- putation and use of mathematical tables.

London where Fisher was professor of Eugenics. It was arranged for the National Machine to be moved to the Galton Laboratory at University College. Comrie's departure was a great loss to the Committee which had been shaped by his leadership and dynamism. This proved unacceptable to the Committee. Comrie had been privately . Comrie took responsibility for supervising most of the computations necessary for the first volume of Bessel Tables using Nautical Almanac Office facilities to do so. was published in 1937. tables of the probability integral. Without Comrie's personal drive and his computing resources the work of the Committee slowed down immensely. Part two was published in 1952. was based on tables donated to the Committee by W. The Mathematical Tables Committee was not the only institution to devote considerable resources to the production of Bessel function tables. Volume 6 of the British Association series.23 he also offered to carry out work on behalf of the Committee at the rate of 4 or 5 shillings per hour. In the 1940s Howard Aiken used the Harvard Mark I to produce 12 volumes of Bessel function tables. The Committee also lost Comrie's practical table making skills and those of his staff. A Bessel Function Subcommittee was created to deal specifically with the preparation of Bessel functions. After Comrie's resigna- tion from the Committee he offered to house the British Association National Accounting Machine at his new office and pay for any use he or his staff made of the machine at a rate of 9 pence an hour.22 Volume 7 of the British Association Series. Sheppard shortly before his death and which the Committee took through to publication. who were con- cerned about the Council of the British Association's attitude towards the machines passing out of the direct control of Committee members24 as well as about the cost differentials. For example. In 1936 Comrie was dismissed from the Nautical Almanac Office for carrying out external computations using Nautical Almanac star? and machines. F. On leaving the Nautical Almanac Office Comrie set up a commercial computing business called the Scientific Computing Service and resigned from the Mathematical Tables Committee. The Mathematical Tables Committee had long been concerned with the tabulation of Bessel functions and Comrie ensured that the Committee remained committed to the project. The first collection of Bessel Function tables. parts three and four were issued in 1960 and 1964 respectively by the Royal Society. T A B L E M A K I N G BY C O M M I T T E E 245 decided to use Cunningham's money to prepare the divisor and power tables for publication and both were issued in 1940.

N. R Miller 1933 G. G.Mordell 1948 H. G. 1877 MVWilkes 1938 A.L. Bickely 1934 H.E. Khabaza 1964 E. Hippisley 1919 A. Milne-Thomson 1939 B.Robbitis 1930 I. H. Smith 1871 W. P. Neville 1929 A. Clcnsliaw 1962 E. H. G. M. Lodge 1888 E Sandon 1938 LordRaylcigh 1888 L. hice 1932 A. L. Henderson 1927 £.J. G. C.S.Olvcr 1959 A.Frwin 1928 H.Massey 1949 G. S. Kuigliding 1958 L.Webster 1912 A. Copal 1949 H. L.W.MacMahon 1895 L.906 EYates 1948 j.Doodsou 1915 Z. H.Whitwdl 1928 C.Jones 1948 M. G. Chaundy 1914 D.R.S. C. CAitken 1948 A.Comric 1928 EJ.E Tocher 1928 M.R. Young 1956 D.R.W.Airey 1911 L. A.J.A. Filon 1906 C.W. Price 1888 R.W. N. Sr.A. M.-Fishe 1925 •L. L.W. Hazelgrove 1957 • R . G.J. Glaisher.LOVC 1 913 L. O.M. Page 1959 J. Members of the Mathematical Tables Committee in date order of joining the committee J.Wishart 1928 D.Jeffreys 1949 G. MacDonald 1913 M.T. Delgarno 1962 F. Fo x • 1958 j. B. Savidge 1915 H. Cashen 1941 W M . Cunningham 1895 E.Watson 1916 H.Hartree 1948 j. Pearson 1930 M. Scorer 1954 R.O. S.W Hobson 1913 C G.Thoiiipsou 1928 E. Nicholson 1908 D. Rosenhead 1948 A. Swinnerton-Dyer 1960 j. M. Stevens 1936 J.Cay!ey 1871 J.Bartlett 1962 T. Glaisher 1871 D. S. Barnett 1965 . C.H. Goodwin 1948 P. A. Newman 1948 E. B.Hill 1. Hicks 1888 J. j.Wommley 1944 A. Howarth 1949 L.W.Kennedy 1916 C. Sadler 1932 W Thomson (Kelvin) 1871 E.T.Todd 1944 A.Matthews 1916 R.Darwin 1949 T.J. M. Gilles 1949 A. Bates 1960 J. Greeiihill 1888 J.Wriiich 1923 C. A. W.J. Coulson 1954 G. Stokes 1871 W.

T A B L E M A K I N G BY C O M M I T T E E 247

employing a computer to assist with the Bessel function compulations but
with Connie's resignation her services were lost. While Committee mem-
bers tried to get extra part-time computing support from colleagues and
students25 the task was becoming increasingly difficult and in January 1937
the Committee approved the employment of a full-time computer. Frank
Cleaver, who had been put forward for the post by the National Cash
Register Company (manufacturers of the National Accounting Machine),
had a starting salary of -f^l WsOd for a 33 hour week and worked in the
computing room of the Gallon Laboratory at University College, London.26
The Committee continued to employ a full-time computer until 1941
when war time labour restrictions meant it was no longer possible. Despite
having access to a full-time computer Committee members, and some vol-
unteers, continued to work on the Committee's behalf.
At the outbreak of the Second World War in 1939 the output of the
Committee dropped dramatically. Many Committee members soon took
up war time posts which meant that they had little or no time for
Mathematical Tables Committee work or meetings. The Committee lost its
full-time computer to war related work and, while Miller at Liverpool
could occasionally employ temporary computers during the University
summer vacations, the situation was far from satisfactory. In addition, the
Government leased the Committee's National Accounting Machine and
moved it to Rothamsted where it was used for war related computations.

Post-war difficulties, 1945-8
The Second World War saw a huge increase in scientific computing and
research into more efficient and effective computing methods. Electronic
computers had their origins in ballistics and code-breaking work while the
armed services sought to increase their computing power as much as they
could using calculating machines, punched card machines, and analogue
computing instruments. Mathematical tables were now only one of a range
of computing tools but interest in them was still strong. For example the
American journal Mathematical tables and other aids to computation began
publication in 1943 and its main aim was to disseminate information about
new published and unpublished mathematical tables.
In Britain, in the immediate post-war period, electronic computers
began to be built at Cambridge University, Manchester University, and the

248 TABLE M A K I N 6 BY C O M M I T T E E

National Physical Laboratory. They promised a huge increase in computing
power to the few people who would have access to them but initially
would be of limited use to the majority of mathematicians, scientists, and
engineers working in the rest of the country. This potential increase in
computing power coupled with an increasing demand for mathematical
tables led some members of the Mathematical Tables Committee to believe
that the late 1940s and 1950s would be a golden era of mathematical table
making. Their expectations were never realized.
By late 1945 members of the Mathematical Tables Committee had
began to think about restarting their activities. Miller in particular had
ambitious plans for the Mathematical Tables Committee and felt that the
Committee should work in partnership with the new Mathematics
Division of the National Physical Laboratory and the Cambridge
Mathematical Laboratory. Both institutions had been set up to provide a
range of computing services to their users and both were involved in build-
ing electronic computers. In Miller's view either or both of these institutions
could be called upon to carry out some of the work of the Mathematical
Tables Committee but that it was essential that the Committee itself remain
separate from them in order that the Committee's role of independently
producing high quality fundamental tables should not be lost. Miller and
William Bickley, a committee member 1934—65, both proposed that the
Committee hire a graduate who could supervise table making activity and
undertake research and at least one ordinary computer at a cost of about
.£500 per annum. Publication and secretarial expenses would need to be
met on top of this. Miller was of the view that the Committee should be
expanding to meet modern needs and not drifting into an advisory role as
had been suggested by some members of the Committee.27
The main obstacle to these plans was finance. After the war the British
Association declared that it would fund research committee work for a
maximum of five years only. The Mathematical Tables Committee was
therefore facing a complete cut in funding. One option was to continue as
a British Association Research Committee and try to exist on government
grants but this would not give the Committee any long term stability.
Another issue of concern at the time was that the British Association
Mathematical Tables Committee had little contact with table users.
Mathematical table users outside the Committee privately questioned the
choices the committee had made in the tables they had produced. In many

TABLE M A K I N G BY C O M M I T T E E 249

cases members had joined the Committee to make tables IB which they had
an interest and found the Mathematical Tables Committee a supportive
place in which to do this. Senior figures in the British Association now
sought to ensure that only tables with practical applications were sponsored
by the Committee. Sir Charles Darwin, Director of the National Physical
Laboratory and a member of the British Association Council, exerted a
strong influence over decisions concerning the Mathematical Tables
Committee's future. He expressed the opinion that the British Association
should no longer be funding table making as the new Mathematics
Division of the National Physical Laboratory (for which Darwin was
directly responsible) could take over much of the computational work.28 In
Darwin's view an advisory body was what was needed. A further influence
was the increasing awareness of computing issues within the Royal Society
which was recommending to universities that computation and computa-
tional methods be top of the list of priority subjects to be taught to post-
war undergraduates.29 In response to these influences the Council of the
British Association recommended that the Royal Society take over from
the British Association as the sponsoring body.
Many members of the Committee fought vigorously against the British
Association's decision to end its commitment to making mathematical
tables, but the Committee was divided between members such as Miller,
Sadler, and Bickley who wanted to push ahead for an active and expand-
ing Mathematical Tables Committee and an older, tired generation of
Committee members who were happy to become an advisory body. Once
a decision had been taken by the Councils of the British Association and
the Royal Society, the Committee membership was left with little say in
the matter. They eventually bowed to the inevitable and entered into nego-
tiations with the Royal Society. In June 1948 the Committee's assets and
liabilities were transferred to the Royal Society and a new Committee
constituted.

The Royal Society Mathematical Tables
Committee 1948-65
The Royal Society Mathematical Tables Committee of June 1948 was a
new Mathematical Tables Committee set up along traditional Royal Society

250 TABLE M A K I N 6 BY C O M M I T T E E

committee lines. Its members, many of whom, were Fellows, were selected by
the Royal Society Council and the secretarial function performed by Royal
Society administrators. The Committee was dominated by senior people in
British computing. Representatives from Manchester, Cambridge, Liverpool,
and the National Physical Laboratory sat on the Committee as did other sen-
ior figures. Less than half of the Committee was made up of members of the
British Association Mathematical Tables Committee. Maurice Wilkes, Director
of the Cambridge Mathematical Laboratory, committee member 1938—48
and chairman 1958—69, expressed the principle difference between the two
committees. He wrote

The [British Association] Committee consisted of a group of people who had long
worked together and shared a similar enthusiasm, moreover, the committee itself
decided on new members. It may thus be said to have had a personality of its own.
The new... committee consisted of individuals arbitrarily selected by the Council of
the [Royal Society] and appointed for limited periods. It had no personality. The take
over made no-one happy. There was little enthusiasm for mathematical tables any-
where in the Royal Society.., The new arrangements were unwelcome to members
of the old committee who would have preferred to have been left alone to get on
with the job as they had always done.30

Wilkes is being a little harsh in these comments in that the Royal Society
Fellows do not appear to have been randomly selected to sit on the
Committee but had all been involved in computing matters of one form
or another. He was, however, correct in his assessment that there was 'little
enthusiasm' within the Royal Society for mathematical tables. The new
devel.opm.ents in electronic computers were exciting 'big' science, on a scale
with nuclear technology, while mathematical tables were simple tools,
From the minutes of the Royal Society Mathematical Tables Committee
and letters between Committee members it is clear that the first year of the
Committee was a difficult one.31 The Committee's terms of reference were to
(i) co-ordinate the activities of those engaged in the calculation of tables
and decide priorities;
(ii) publish fundamental tables of the highest standards of accuracy and
presentation;
(iii) give advice on the mathematical and computational aspects of tabula-
tion and on the arrangement and presentation of tables;
(iv) administer special funds for the preparation of tables, e.g. the
Cunningham Bequest.32

TABLE MAKING BY COMMITTEE 251

Initially the actual work of producing and tabulating mathematical tables
was to be carried out by a General Sub-Committee which, in the main,
drew its members from the British Association Committee. The role of the
General Sub-Committee was to deal with actual table making decisions but
it had no executive powers and could not initiate any work without first
seeking approval from the main Committee, Members of the General
Sub-Committee fought against their subordinate role, complained bitterly
about the lack of progress being made and about the unrepresentative min-
utes being produced by Royal Society secretaries.33 In May 1949 the main
Mathematical Tables Committee bowed to pressure and the General Sub-
Committee was dropped; the most active members of the British
Association, Committee taking places on the main Committee. In May
1949 Comrie was invited to serve on the Royal Society Mathematical
Tables Committee but by this time he was ill and able to act only in a lim-
ited advisory capacity.
As things settled down the Royal Society Mathematical Tables Committee
began to work on a series of mathematical tables many of which had their
origins in the programme of work transferred from the British Association
Committee. But progress was slow despite the Committee hiring a full-time
computer from December 1949. The first volume of the Royal Society

Fig. 9,4 Mathematical Table
Committee sates figures. A table
showing the 1951 sales figures for
published British Association and
Royal Society Mathematical Tables
Committee volumes. While this
represents only a snap shot of the
sales figures it does illustrate how low
the number of tables sold was. (from
Royal Greenwich Observatory Archives,
Cambridge University Library, RGO 16
100C Folder VI by permission of the
Syndics of Cambridge University
Library and of the Particle Physics and
Astronomy Research Council.)

252 TABLE M A K I N 6 BY C O M M I T T E E

Mathematical Tables Series, The Farey series of order 1025, was published in
1950, Volume 2 was issued in 1956, Volume 3 in 1954, and Volume 4 in
1958, There were several reasons why the production, rate was so slow. First
many of the Committee's principle table makers, Miller, Sadler, and
E.T, Goodwin who joined the Royal Society Committee in May 1950 as a
representative from the National Physical Laboratory Mathematics Division,
were becoming extremely busy in their working lives and had only a lim-
ited amount of time to give to table making; all were involved in some way
with the application of electronic computers. Secondly, and perhaps most
importantly, the Royal Society Mathematical Tables Committee suffered
from a lack of enthusiastic leadership and from having too many policy
makers on the Committee and not enough workers.
While Wilkes observed that there was little enthusiasm for mathematical
tables in the Royal Society; this was also true of many of the senior mem-
bers of the Mathematical Tables Committee. For example, when Charles
Darwin was chairman of the Royal Society Mathematical Tables
Committee 1951—6 he would repeatedly question the value of the tables
being computed which tended to have a dampening effect on the work.34
The lack of activity was obvious also from the few meetings held (some-
times down to one or two a year) and the lack of business waiting to be
executed. This feeling was summarized in 1953 by Alexander Thompson,
a Mathematical Tables Committee member since 1929, who wrote *So long
as the management of a Committee has no real interest in the work,
progress is bound to be slow and uncertain'.35
Despite the slow progress being made on preparing and publishing tables
the Royal Society Mathematical Tables Committee developed other roles
to serve the computing community. One such role was the development of
a depository of unpublished tables. In January 1950 the Mathematical
Tables Committee agreed in principle to collect unpublished mathematical
tables and deposit them in the Royal Society Library,3ft The intention was
to collect together those tables offered to the Committee but which proved
to be too small or too specialized to realistically publish. The idea that the
Committee should be actively seeking tables for deposit soon came to the
fore and the depositary was formally announced in early 1951 in the
journals Nature and Mathematical tables and other aids to computation37 And
anyone was invited to submit tables for deposit or use the depository. By
late 1953 forty two tables had been accepted into the depository and the
collection continued to grow so that by 1965 it held 86 tables.

TABLE M A K I N G BY C O M M I T T E E 253

Short biographies of selected members of the
Mathematical Tables Committee
J.W.L. Glaisher (1848-1928)
James Whitbread Lee Glaisher was bom in 1848 the sou of James Glaisher
astronomer, mathematician, and meteorologist. He weut to Trinity College,
Cambridge graduating second wrangler in 1871 when he was elected a Fellow
of Trinity College. Glaisher spent the rest of his working lite writing and lec-
turing in Cambridge publishing widely in many branches of mathematics
including over 40 articles OB mathematical tables. He was editor of die Quarterly
journal of mathematics and the Messenger of mathematics for many years. Glaisher
was aw active member of the British Association, the Royal Astronomical
Society, and the London Mathematical Society. He was elected Fellow of the
Royal Society in 1878. Glaisher served on the Mathematical Tables Committee
from 1871 to 1901 serving as the Committee's first Secretary.

James Glaisher (1809-1903)
James Glaisher began his career working on the ordnance survey of Ireland
and was appointed First Assistant at the Cambridge Observatory in 1833 by
George Airy. On Airy's appointment as Astronomer Royal, James Glaisher
moved with him to Greenwich. Under Airy, James Glaisher gained extensive
experience in mathematical computation and table making. In 1840 James
Glaisher was appointed Superintendent of die Magnetic and Meteorological
Department at Greenwich. He held the post for 40 years gaining a reputation
as a pioneer in meteorology. He was also a well known balloonist. James
Glaisher served on the Mathematical Tables Committee from 1877 to 1889.

Arthur Cayley (1821-1895)
In 1842 Arthur Cayley graduated from Cambridge as senior wrangler and
soon began to publish original papers. In 1863, after a successful career in the
law, Cayley was appointed Sadlerian professor of mathematics at Cambridge.
Cayley published widely in many branches of pure mathematics and pub-
lished several mathematical tables. Cayley was elected FRS in 1852 and was
a very active member of the British Association. Cayley served on the
Mathematical Tables Committee from 1871 to 1895.

Alfred Lodge (1854-1937)
After graduating from Magdalen College, Oxford, Lodge was appointed
Professor of mathematics at the Royal Indian Engineering College. In 1904
he took up the post of assistant mathematical master at Charterhouse School
where he stayed until his retirement in 1919. Lodge was the first President of

Short biographies of selected members of the
Mathematical Tabies Committee cont
the Mathematical Association and Secretary of Section A (Mathematics and
Physics) of the British Association from 1888 to 1892. In addition to teaching,
Lodge was a tireless and extremely competent computer. He published nine
mathematical tables and was a member of the Mathematical Tables Committee
from 1888 to 1937 serving as Secretary from 1888 to 1896.

Allan Cunningham (1842-1928)
Allan Joseph Campneys Cunningham was born in India, educated in England
and was commissioned into the Bengal Engineers in 1860 in which he both
taught mathematics and saw active service. Cunningham retired from the
Army as Lieutenant Colonel in 1891 and taught mathematics at die School
of Military Engineering at Chatham from 1895 to 1900. Following his retire-
ment from the Army Cunningham published extensively on mathematical
tables. On his death in 1928 he left legacies to the British Association and the
London Mathematical Society to further his work on number theory tables.
Cunningham, was a member of the Mathematical Tables Committee from
1895 to 1901 serving as Secretary from 1896.

John Airey (1868-1937)
Following a 1894 degree from London University, John Airey taught first at
Forth County School in Glamorganshire before going to Cambridge where
he took a second degree in 1906. Airey was principal of Merely Secondary
School 1906-12, principal of West Ham Technical Institute 1912-18, and
then principal of the City of Leeds Training College until his retirement in
1933 when he became a co-editor of Philosophical magazine. Throughout his
life Airey was a prodigious calculator and table maker with a special interest
in Bessel functions publishing over 40 mathematical tables during his lifetime.
Airey was a member of the Mathematical Tables Committee from 1911 to
. 1937 .serving as Secretary from 1920 to 1929.

William Bickley (1893-1969)
In 1913 William Gee Bickley graduated from University College, Reading
and began a career teaching mathematics and engineering. In 1930 he was
appointed Reader and Assistant Professor at Imperial College, London teach-
ing mathematics to engineering students. In 1947 Bickley was appointed
Professor of Mathematics at Imperial, a post he held until his retirement in
1953. Despite failing eyesight, Bickley computed and published mathematical
tables during the late 1930s and 1940s. He went completely blind in 1949 but

continued to lecture and write despite his disability. Bickely was a member of
the Mathematical Tables Committee from 1934 to 1965.

L.J. Comrie (1893-1950)
Before serving in the New Zealand Expeditionary Forces during the First
World War Leslie John Comrie graduated from University College, Auckland.
After the war Comrie took a Ph.D. at Cambridge and taught in the United
States before joining the Nautical Almanac Office in 1925 where he was soon
appointed Deputy and then later Superintendent While at the Nautical
Almanac Office Comrie built up a reputation as Britain's leading mathemati-
cal table maker and expert oil machine computation. Comrie left the
Nautical Almanac Office in 1936 and set up the Scientific Computing
Service, the first commercial scientific computing bureau of its kind. Comrie
was a member of the Mathematical Tables Committee from 1928 to 1936 and
from 1949 to 1950 serving as Secretary 1929 to 1936.
J. C. R Miller (1906-1981)
Jeffrey Charles Percy Miller studied both mathematics and astronomy at
Cambridge under A. S: Eddington before becoming a research 'assistant at
Imperial College. In 1935 he became a lecturer at Liverpool University before
joining Conine's Scientific Computing Service as technical director from 1946
to 1950. In 1950 he was appointed to the Cambridge Mathematical
Laboratory. Miller's main interests were mathematical tables and numerical
analysis and he published widely. Miller was a member of the Mathematical
Tables Committee from 1933 to 1965 serving as Secretary from 1946 to 1948.
D. H. Sadler (1908-198?)
Donald Sadler joined the Nautical Almanac Office in 1930 after graduating
in mathematics at Cambridge earlier that year. Sadler was trained in math-
ematical table making and numerical computing by Comrie and he became
first Deputy and, in 1936, Superintendent in Comrie's footsteps. Sadler
remained Superintendent of the Nautical Almanac Office until his retirement
in 1971. During the -war he organized computations for the Admiralty
Computing Service and served as Cieneral Secretary of the International
Astronomical Union 1958 to 1964. Sadler was a member of the Mathematical
Tables Committee from 1932 to 1965.
E.T.Goodwin (1913-)
E. X Goodwin, 'Charles' to his friends, graduated from Cambridge in math-
ematics in 1934 and gained a Ph.D. in 1938. During die Second World War he
worked for the Ministry of Supply and the Admiralty on a wide range of math-
ematical problems. In 1945 he joined the newly created Mathematics Division
of the National Physical Laboratory being promoted to Superintendent in 1951.

256 TABLE M A K I N 6 BY COMMITTEE

Short biographies of selected members of the
Mathematical Tables Committee con/.
He retired in 1974 as Deputy Director (C) of the National Physical Laboratory'.
Goodwin sewed as a member of the Mathematical Tables Committee in 1948
and from 1950 to 1965.

M.V.Wilfces (1913-)
Maurice Wilkes took a degree in mathematics at Cambridge in 1934 and a
Ph.D. in 1938. In 1937 he was appointed as a demonstrator in die newly cre-
ated Mathematical Laboratory at Cambridge. After war service Wilkes was
appointed Director of the Cambridge Mathematical Laboratory where he
designed and built EDSAC, one of die world's first electronic computers. Since
his 1980 retirement from the renamed Cambridge Computer Laboratory,
Wilkes has held senior consulting positions with the Digital Equipment
Corporation, Olivetti and AT&T. He was elected FRS in 1956. Wilkes was a
member the Mathematical Tables Committee from 1938 to 1948 and from
1958 to 1965 serving as Chairman for the latter period.

The tables came from a variety of sources. Tables made by government
research establishments such as the National Physical Laboratory and the
Admiralty Research Laboratory made up some of the collection. The
Admiralty also deposited tables which had been made by the Admiralty
Computing Service during the war. Some tables were submitted by
members of the Committee and others came from industry and from
universities. From time to time the Committee published lists of the tables
held in journals such as Mathematical tables and other aids to computation,
Philosophical magazine^ and the Journal of the London Mathematical Society, and
encouraged people to use the facility in person or request copies. Users of
the tables are harder to locate. A list of early users issued in 195438 shows
that institutions as diverse as the National Research Council of Canada,
North American Aviation Inc., the US Navy, Ferranti Ltd, and the British
Broadcasting Corporation all used the service.
Another role played by the Royal Society Mathematical Tables
Committee was as a forum, for technical debate on different aspects of table
making or computation techniques. The most interesting of these discus-
sions concerned the effect of the increasing number and sophistication of
electronic computers on the production of mathematical tables.

TABLE M A K I N G BY COMMITTEE 257

The advent of electronic computers, 1952-65
In 1952 the Royal Society Mathematical Tables Committee held the first
of its discussions on the effect of electronic computers on mathematical
tables. It was prompted by an invitation from John Todd for a member of
the Royal Society Mathematical Tables Committee to attend a symposium
on Mathematical tables in the age of electronic computers to be held in
Washington, DC. Todd, a member of the British Association Mathematical
Tables Committee 1944—8, now ran the Computation Laboratory of the
National Bureau of Standards in Washington. Todd was hosting the sympo-
sium, held on 15 May 1952, in response to suggestions that once 'high
speed digital computing machines are readily available, the need for math-
ematical tables, in the conventional form, will disappear'.39Todd hoped that
by bringing together as many table makers and table users as possible that
it would be possible to outline a long term policy for the future needs of
the table using community. Before the symposium Goodwin and Miller, in
consultation with other active table makers on the Committee, reported on
the matter to the Mathematical Tables Committee. In their view there were
three stages to mathematical table production: planning, computation, and
publication. Electronic computers, when they became widespread, would
ease the burden of the computation stage by a factor of about 100 but the
amount of planning that went into producing a table had remained fairly
constant. The greatest part of the labour was now publication which was
both very expensive and time consuming.
The Committee agreed to send Goodwin to the symposium and in
preparation he canvassed opinions from other interested parties in Britain.
The consensus of opinion, expressed in a Mathematical Tables Committee
report,40 was that the very small number of computers which were likely
to be available in the foreseeable future was unlikely to meet the comput-
ing needs of the majority. In the Committee's opinion tables would con-
tinue to be needed and the number of machines available in England was
too low to impact significantly on the computing of tables. The whole tone
of the report is slightly dismissive of the notion that electronic computers
would have a big impact on either the need for tables or their production.
It differs subtly from Goodwin and Miller's earlier note and illustrates
differences of understanding between those working at institutions
which already had operational computers, such as Miller at Cambridge and
Goodwin at the National Physical Laboratory, and those with less direct

258 T A B L E M A K I N 6 BY C O M M I T T E E

involvement. The following quote taken from, the opening paragraph of the
report illustrates the point.
The need for extensive tables to moderate accuracy will continue especially for use
in small organizations and individual workers. Tables of certain types of function
(eg. those which machines would take a long time to compute) will be required to
higher accuracy and the need will increase. The number of machines expected to
be in use in the U.S.A. within the next year (50) is many times too small if tables
were to be dispensed with.Tables are made by man and not machine.41
Goodwin attended the symposium along with C.W.Jones, who served on
the Bessel Function Panel of the Mathematical Tables Committee from 1947
onwards and was a member of the full Committee 1948 and 1953—61.
Goodwin had felt apprehensive about going to the Washington meeting as he
felt their ideas were vague and that planning for the unknown was difficult
but he need not have worried. The situation in America turned out to be
even less clear than in Britain. Much of the meeting was taken up with dis-
cussions concerning the types of tables and problems which could be applied
to electronic computers. The need to continue making tables to serve the
thousands of scientists who would not have access to computers was dis-
cussed but the closure of the New York Mathematical Tables Project42 in the
late 1940s meant that there were few outlets for mathematical table publica-
tion in the United States and that this was felt to be a growing problem. No
overall conclusions seem to have been reached and Goodwin felt that the
symposium was 'premature*43 and that it was too early to discuss fully what
effect computers would have on the production of mathematical tables.
The debate about the future of mathematical tables was reopened in
1954 when the Massachusetts Institute of Technology hosted a'Conference
on Mathematical Tables: their publication and distribution, together with a
consideration of their use in the light of the advent of high speed comput-
ing machines'.The Royal Society Mathematical Tables Committee did not
send an official representative but Miller attended the conference in his
own right.The driving force behind the conference seems to have been the
difficulty table makers were experiencing getting tables published and only
one session of the conference actually considered if tables still needed to be
made; the answer was a resounding yes.
In 1958 Maurice Wilkes, Director of the Cambridge Mathematical
Laboratory, was appointed Chairman of the Royal Society Mathematical Tables
Committee. His appointment came at a time when it had become obvious to
many that the use of electronic computers would become widespread. Wilkes
was familiar with aE kinds of computing techniques and had served on the

(Photograph by kind permission of Sir Maurice Wilkes. the Committee would have had to reconsider its role in response to that fact. TABLE MAKING BY COMMITTEE 259 Fig. 9. which he took up rather reluctantly under pressure from Royal Society administrators. the final chairman of the Mathematical Tables Committee. Wilkes realized that 'some of the members were still living in the past. had the melancholy job of winding down the Committee's activities in response to the widespread development of electronic computers and calculators which heralded the end of the mathematical table era. He also suggested that the Committee consider tables for use with computers as a possible way forward. Wilkes wrote 'Much as 1 admire the work done by the Committee in the past I cannot feel enthusias- tic about embarking on a future programme of tabulating functions of which individual values can be obtained by a digital computer in a few milliseconds*.) British Association Mathematical Tables Com. On his appointment.4* Wilkes invited comments on his report and over the next year .5 Maurice Witkes. or at any rate had not appreciated the change which had to come about'44 and from the start Wilkes felt his job was to bring the Mathematical Tables Committee to an end. On taking over as chairman Wilkes* first task was to initiate a discussion of the future of mathematical tables on October 1958.mittee 1938—48. He pointed out that the good work of the Mathematical Tables Committee and others had ensured that most of the fun- damental tables likely to be needed by desk machine operators had already been published and stated that even if electronic computers had not been developed. He took on this role in two ways. But Wilkes' true feelings can be detected in one sentence buried in the middle of the report. He produced a report45 in which he stated clearly the changing situation. He noted that electronic com- puters had changed computational needs as well as making the computation of mathematical tables much easier.

1948. 1948. 333—40. 341—7 and reprinted in Mathematical tables and other aids to computation. Early scientific computing in Britain. Croarken. R. For the importance of mathematical tables in Victorian Britain see A.Warwick. 1948—9. The British Association for the Advance- ment of Science: A retrospective 1831—1931. The changing face of scientific computing in the UK over the period 1925 to 1955 is described in M. vol. 1931. 3. Biographical information on some table makers can be found in R. Further reading A history of the British Association for the Advancement of Science up to 1931 is given in O. The Committee was finally dissolved in 1965 and replaced by a skeleton Committee set up to administer the sales income from existing Mathematical Tables Series publications and make reprinting decisions.260 TABLE M A K I N 6 BY C O M M I T T E E Committee members responded. 311-51 in M. Mathematical table makers. and obvious. 1995. Most of the responses suggest that the Committee look towards producing tables for use with computers such as regu- larly spaced values of functions with subroutines for interpolation. Oxford.The Scripta Mathernatica Studies Number Three. To this end seven of the eleven volumes of Royal Society Mathematical Tables Series were published after Wilkes took over the chairmanship of the Committee. Over the next few years Wilkes applied a gentle but persistent pressure to Committee members to either finish or abandon table making projects which had been on-going for several years. British Association for the Advancement of Science. Yeshiva University.Archibald. It was slowly becoming inevitable. C. It was the end of the mathematical table era.Wise (ed). 1990. The reports of the Royal Society Mathematical Tables Committee are held at . that there was no longer any justification for producing traditional mathematical tables. See also the Dictionary oj national biography for more biographical details of some the table makers mentioned. Howarth. London. While many older members of the Committee did not like the fact they did acknow- ledge it. Clarendon Press. vol. The reports of the British Association Mathematical Tables Committee can be found in intermittent volumes of Report of the British Association for the Advancement of Science from 1873 to 1939. N. 5.'The laboratory of theory' pp. The final report is published in The Advancement of Science. Princeton University Press. J. Princeton. The values of precision. New York.

. Notes 1. Cunningham. Report of the British Association for the Advancement of Science 1870. . exponential sine and cosine integrals. 1883. Preface. Report of the British Association for the Advancement of Science 1873. Cambridge. "Tables of the Exponential Function". Taylor & Francis. 'Beautiful Numbers: The Rise and Decline of the British Association Mathematical Tables Committee. Report of the British Association for the Advancement of Science 1873. Cambridge and London. Taylor & Francis. integrals of probability integral. Croarken and M. 1990. 'Report of the Committee. Ixix. London. Report of the British Association for the Advancement oj Science 1873. J.Taylor & Francis. L. Taylor & Francis. J. Tables t)fjunctions. 3 (1878). Jahnke and F Emde. 1871—1965'. 13. 8. 1. J. p. 22-46. 1933. London. 46—57. Factor table for the sixth million. A. See M. 7...net/hniit> accessed November 200L 2. 99-138. Croarken. Report of the British Association for the Advancement of Science 1879. 172. IEEE Annals of the history of computing. pp. 44-61. Deighton Bell. 'Report of the Committee. Currency conversion data taken from <http://eh. A. on Mathematical Tables'. British Association for the Advancement of Science. Ixxi.0. 'Report of the Committee. Clarendon Press.. See also M. 22:4 (2000). See. 17. 6.. Glaisher.. Oxford.. 3. 14. Campbell-Kelly. Factor table for the fourth million. Report of the British Association for the Advamement of Science 1889. p. '[lie theory of sound. in E. MathematicalTables Volume 1: Circular and hyperbolic functions. T A B L E M A K I N G BY C O M M I T T E E 261 the Royal Society and in Sadler's papers held at Cambridge University Library RGO 16 tOOD. 4.on Mathematical Tables'. 1931. A binary canon. for example. Office of the British Association. Glaisher 'On factor tables. 1877 and 1878. Early scientific computing in Britain.. p. Cayley. London. 12. London. Glaisher. *. 'Report of the Committee. 'First Report of the Committee. 243-372.. For more information about Sadler's career see Chapter 11.1—175. 1879. W. factorial (gamma) and derived functions. A treatise on elliptic functions. 28-9. 9. Lord Rayleigh. London. 1876. p. 11. Transactions of the Cambridge Philosophical Society. Proceedings of the Cambridge Philosophical Society. 13 (1883). London. 2.011 Mathematical Tables'. Factor table for the fifth mt7/i'on. 5. 15.. Miller suffered a serious illness in the late 1920s the after effects of which precluded him from . 16. . 1900.W L. Report of the British Association for the Advancement of Science 1871. 1880. 1-175. on Mathematical Tables'.*. Leipzig. 2 volumes.

Hardy and A. Minutes of the Mathematical Tables Committee 5 November 1936. His letters are preserved in the Bodleian Library Oxford. British Association Mathematical Tables Series Vol. Minutes of the British Association Council 46/108. Cambridge 1933. 27. British Association Mathematical Tables Series. 31. 25.262 TABLE M A K I N 6 BY COMMITTEE passing the civil service medical examination which he would have had to have done to gain employment at the Nautical Almanac Office. 1999. Harvard University Press. d. Bodleian Library Dep BAAS 29 folio 283. 3. Bodleian Library. Dickson. British Association Mathematical Tables Series Vol. 1158. C.935. 28. Also students and facilities at the University of Liverpool were used (Report of the Mathematical Tables Committee. 10 November 1936.Emai to Martin Campbell-Kely and Mary Croarken 3 August 1999. Cambridge. Letter from W H.Minutes of the Mathematical Tables Committee 5 November 1936. RGO 16 100D. Report of the British Association for the Advancement of Science 1937. Chapter 8. Sadler 22nd December 1945. E. L. 20. E. Bodleian Library Ms. Misc. Letter from J.'Inverse interpolation and scientific applications of the National Accounting Machine*.V. Cambridge University Library RGO 16 92/4 folder HI. Western among others. Eng. Howard Aiken: portrait of a computer pioneer. d. 18. 1158. The minutes of the Royal Society Mathematical Tables Committee are held at the Royal Society in London. Comrie consulted G. Sadler's papers at Cambridge University Library contain copies of the minutes and letters between Committee members. Factor table giving the complete decomposi- tion of all numbers less than 100000. H. Comrie. Ms. Minute 7. Bickley to D. Eng. Minutes of the British Association for the Advancement of Science Council 7 June 1946. B. A. These probably came from his contacts in the General Register Office. 26. L.net/hmit> accessed November 2001. 29. 24. Supplement to ike journal of the Royal Statistical Society. Misc. . L. Connie). Ms. Minimum decompositions into fifth powers. 3:2 (1936). H. 19. 272—3). 677 Dl—D5. Cambridge. J.Wilkes. 1.. Ms. 4. 23. Cohen. Bodleian Library Ms. H. 22. 87-114. Cambridge University Library RGO 16 92/4 Folder II. Mass. Cambridge Univer- sity Library RGO 16 92/4 Folder II. Minutes of the Royal Society Mathematical Tables Committee 7 November 1947. Ince. Cambridge 1934. Minutes of the Mathematical Tables Committee. Eng.Vol. 30. d.J. d. Sadler 16 December 1945. Eng. Data taken from <http://eh.J. E Miller to D. 32. Misc. 5. 21 January 1937.Thompson could supply computing labour from time to time but not on a regular basis. I. Eng. E. L. Misc. 1158. Bodleian Library. M. Cycles of reduced ideals in quadratics fields (ed. 1158. Approximately equivalent to _£ 135 GOO today. Minutes of the Mathematical Tables Committee. 21. Misc.

46. Royal Society Mathematical Tables Committee Report MX/15(58). Miller to D. Mathematical tables and other aids to computation 5 (1951). Cambridge University Library RGO 16 100C Folder IX. See David Grier (Chapter 10) on the New York Mathematical Tables Project. G. P. Cambridge University Library RGO 16 100C FolderVI. 1. M. p. 37. Sadler to J. Royal Society Mathematical Tables Committee Report MX/15(58). Cambridge University Library RGO 16 100C Folder VI. Typescript report Cambridge University Library RGO 16 100C Folder VI. P. W.V. manuscript note dated 12 December 1978 attached to Royal Society Mathematical Tables Committee Correspondence file in Wilkes' possession. Wilkes. 'A conference on mathematical tables in the light of electronic computers'. 'Symposium on "Mathematical Tables in the Age of Electronic Computers" *. D. Nature. 41. Wilkes 'The future of mathematical tables'. Thompson to D. 303. Permission to quote has been kindly granted by the Syndics of Cambridge University and the Particle Physics and Astronomy Research Council. M. 83—4. 'Unpublished mathematical tables'. Permission to quote has been kindly granted by the Syndics of Cambridge University and the Particle Physics and Astronomy Research Council. 'NBS Symposium on mathematical tables in the age of electronic computers'. Cambridge University Library RGO 16 100C Folder XIII. Sadler 20 February 1. H. H. 'Tables supplied from the depository of unpublished mathematical tables (up to no. 'NBS Symposium on mathematical tables in the age of electronic computers'. p. Sadler 2 July 1948.Wilkes. Cambridge University Library RGO 16 1OOC Folder XIII. 42. Cambridge University Library RGO 16 100D Folder I. . V.V. Letter from A. Letter J. 35.Wilkes 'The future of mathematical tables'. 44. J. 34. Vol. 'Royal Society's depository of unpublished mathematical tables'. 24 February 1951. minute 5. Cambridge University Library RGO 16 100C FolderV Permission to quote has been kindly granted by the Syndics of Cambridge University and the Particle Physics and Astronomy Research Council. 40. C. Email to Martin Campbell-Kdly and Mary Croarken 3 August 1999. M. Permission to quote has been kindly granted by the Syndics of Cambridge University and the Particle Physics and Astronomy Research Council. Minutes Royal Society Mathematical Tables Committee 18 January 1950. Brunt \ February 1949.l. Bickley to D. 36. p. 20)'. Cambridge University Library RGO 16 100C Folder VII. Miller 8 July 1948.953. M. 45. 38. C. Royal Society Mathematical Tables Committee Report MX/15(52)(Amended). Royal Society Mathematical Tables Committee Report MT/10(54). Royal Society Mathematical Tables Committee Report MT/12{52). Royal Society Mathematical Tables Committee Report MX/15(52)(Amended). H. 43. TABLE M A K I N G BY C O M M I T T E E 263 33. 167. 39.V.

.

'We were modelled after them. its leaders liked to compare their efforts to those of Gaspard Riche de Prony (described by Ivor Grattan- Guinness in Chapter 4). he would claim.1 Tables of probability functions published in 1941 was a typical example of the smaller interval high decimal figure tables which the WPA Mathematical Table Project published. Herbert Salzer. the best evidence suggests that the founders of the Mathematical Tables Project originally knew noth- ing about de Prony and his cadastral computers. . who was one of the assistant leaders of the group. who served as Fig. 10 Table making for the relief of labour DAVID ALAN GRIER After the Mathematical Tables Project had shut its doors in New York and consigned its worksheets to the flames.'We were like those French mathematicians'. would grow animated when he talked about the two organiza- tions. 10. Salzer.'1 In spite of his protestations.

their human computers had no pro- fessional standing. as Daniels suggests. The leaders of the project. Scientists first assemble a body of knowledge and practices. Arnold Lowan and Gertrude Blanch. and that the final values could be trusted. Next. the group could count itself a success. they produced a popular set of small 'Lilliputian' tables and hundreds of specialty computations for the Army. None knew advanced mathematics.266 T A B L E M A K I N G FOR THE RELIEF OF L A B O U R bibliographer for the group. Though their major product. Both faced sceptical publics that were reluctant to accept the fruits of their work. private corporations. Few knew any alge- bra. in his study of the professionaliza- tion of American science. though with telling differences that underscore the extent to which computation has its roots not only in the intellectual dis- cipline of mathematics but in the physical discipline of labour. was at best lightly used. they establish the legitimacy of their organization in the eyes of the public. For the most part. but among scientists. and university professors. probably made the connection while searching for old mathematical tables in the New York Public Library or the New York Engineering Society. attempted to give their workers a pro- fessional appearance by imposing a rigorous. the National Bureau of Standards. When the project began operations in 1937. Such struggles are nothing new to science and are commonly found among scientific organizations that are trying to build a professional identity.2 The Mathematical Tables Project had to establish its legitimacy not among the general public. George Daniels. that their computers were adequately trained. they suffered the damning criticism of faint praise from scientists. even if he created the connection after the fact. The Mathematical Tables Project proved the more successful of the two computing organizations. they build an institution to support that knowledge. Yet. nearly mechanical form of . If their legacy had been nothing more than the Navy's LORAN navigation tables or Han Bethe's thermodynamic calculations. the Navy. Both employed large numbers of relatively unskilled computers. the Mathematical Tables Project was con- stantly forced to argue for the value of its work. Most understood only the rules of addition and even these didn't fully apprehend all the properties of basic arithmetic. Both were born in times of social travail. has argued that the process of building a profes- sional discipline involves three stages. The leaders of the Mathematical Tables Project fallowed much the same pattern. Finally. twenty-eight volumes of mathematical tables. Over its ten year existence. Salzer was correct in drawing a parallel between the two groups. Each new calculation required the project leaders to claim that they knew what they were doing.

They even welcomed strict peer review of their own activities. The National Academy of Sciences was located just a few blocks from Morrow's WPA office. told them of the proposed Mathematical Tables Project and asked if they might be willing to appoint a scientific advisory committee. Morrow's first task was to find a sponsor for the project. Lowan and Blanch were able to release excess workers. It was housed in a white marble building that appeared to be more of a Masonic temple to science than the home of an advisory body. which was initially interested until university officials discovered that they would be required to provide twenty per cent of the projects budget. Unable to find another sponsor from the academic community. financing graduate fellowships. They wrote strict instructions and prepared worksheets designed to identify mistakes. an associate statistician in the Washington WPA office. The officers. he asked the New York mayor's office to sponsor the group. who were represented by their permanent secretary. adopt modern computing machin- ery. As the Second World War engulfed the United States. When they agreed. both accidental and deliberate. The president of the academy wrote to the WPA. and sponsoring meetings. The organization was formed during the civil war as a means of recog- nizing successful scientists and as a way to coordinate scientific advice for government projects.3 It was organized in a hurried manner by Malcolm Morrow.' Furthermore. In the end. Origins of the Mathematical Tables Project The Mathematical Tables Project was created by the Work Projects Administration (WPA). and handbooks. 'the National Academy of Sciences in its part would be glad to appoint a committee of . an organization that would take responsibility for daily operations. He approached New York University. bibliographies.'I wish to assure you that we shall be most happy to render any service within our power. Morrow approached the leaders of the academy. the group was truly successful only when it was able to shed its work relief origins. and distance themselves from the problems of work relief. a government agency that attempted to create jobs for 'employable workers'. were generally pleased with the proposal. he turned to the National Academy of Sciences to provide the necessary mathematical expert- ise. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 267 discipline upon the group.The academy also attempted to shape the scientific com- munity by publishing monographs.They organized the computers by arithmetical opera- tion and gave them Ettle freedom in their work.

WPA workers constructed a new statistics building for the campus. Many dealt with. to develop the resources of this key area. in order to prevent a run on the money supply. writer's and theaters projects and the historical records survey. The WPA undertook three kinds of projects. dams. They paid Iowa State University to conduct a large survey of rural life and. the first project.The most prevalent were con- struction projects. art. they paid the salary of many computers.* This project hired writers to prepare guidebooks for states. the work of a single scientist or laboratory. Congress terminated funds for theatrical productions in 1939. school lunch programmes. airports. Originally called the Work Progress Administration. parks. social or economic research. employed workers to review local government records and create historical indices. and organized library card catalogues. During his first hun- dred days in office. which would later be known as the New Deal. The WPA funded about 4000 science projects.This was a 'single nation-wide project which. The Work Projects Administration (WPA) was the third New Deal public works program. garden- ing. sewing and knitting projects. the WPA sponsored cultural activities under a program known as Federal Project Number 1 . it was formed in 1935 to operate 'a nation-wide program of "small useful projects" designed to provide employment for needy employable workers*. They also funded a computing group in Philadelphia to do map computations for the Coast and Geodetic Survey.This group also sponsored the Federal Theatre Project. in the process. and canning efforts. Finally. They included the Agricultural Adjustment Administration.' The project was criticized by conservatives in the press and the American congress. It built on the lessons of its predecessors by providing funds to local agencies who would then Use the'moneys to construct public works. his administration would initiate a collection of programs. He then set his new administration to implement a series of relief programs that would employ those who had lost their jobs and would rebuild the economy.. which required them to support labour intensive activities. and government offices.The WPA built roads. His first act was to declare a bank holiday. Equally prevalent were service projects: adult education classes. Work Projects Administration When Franklin Roosevelt became President of the United States in March 1933. provided a central administration for music. nurs- ery schools. Orson Welles' play Cradle will rock particularly drew public anger. a group that WPA officials hoped would 'lead to the establishment of muni- cipal theatres in a number of cities and even a nationally endowed theatre. with WPA sponsorship. the National Recovery Administration. he took the leadership of a country whose economy had been con- tracting for nearly three and one half years. for relief of industry. Most of them were small activities. and die Tennessee Valley Authority. . By the nature of their charter. diough most were engaged in statistical tabulation.

Most importantly. (Photograph courtesy of the Library of Congress. Before he departed. 10. a former president of the American Mathematical Society who had led the Army's mathematical ballistics efforts during the First World War. and lesser known scientists who were considered skilled table makers. Residents of New York City queuing for food underneath the Brooklyn Bridge. They concluded that the chair of the committee should be Oswald Veblen. a leading example being Harold Davis of Indiana University. such as Vannevar Bush of MIT. he needed to find trained mathematicians to lead the group.2 A New York Breadline.'5 His experience in relief work had taught him the perils of announcing a project before it was ready.'4 With this promise of advice for his project. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 269 Fig.) mathematicians to advise on specific undertakings. They considered well-regarded scient- ists. they had . the leaders of the academy dis- cussed possible committee members.6 By the time that they had decided upon Veblen. circa 1936. he asked the officers of the academy that 'no intimation of consideration of the project be permitted to be made publicly or find its way into the channels of the press. He still needed to file a formal application for the project and find adequate office space. Waiting patiently for Morrow to return. Morrow left for New York.

Morrow was in here saying we were to get a letter the next day'. had been one ofVeblen's assistants during the war.'8 Starting operations When Malcolm Morrow proposed the project to the National Academy. Morrow had changed his plans without inform- ing the academy. The WPA recruited two underemployed scientists to direct the project. .' He said that the WPA could pay the salaries of the directors. 'Mr. He seems to have disappeared off the horizon. The budget proved to be a small fraction of what Morrow suggested. the Committee on the Bibliography for Mathematical Tables and Other Aids to Computation (MTAC). 'I have wondered what has happened to him.The offices were an old industrial loft and the only computing machines for the groups were those that had been abandoned by other government agencies. He even specul- ated that the project could get an extra grant of $100000 for the purchase of computing machines. the director of the National Bureau of Standards. Lyman Briggs. He mentioned that the white collar relief budget for New York City was $2300000.*7 In fact. Briggs took immediate charge of the project. A. he recruited an existing committee from the National Academy of Sciences. The leaders of the National Academy of Sciences reacted politely when they learned of the meeting.' wrote the chair of the group. would provide furniture and office machines. Bennett of Brown University. 'that the Committee work should be limited to a bibliography. wrote one of the academy staff members. The scientists who oversaw the MTAC committee felt that their authority had been circum- vented. other members of the academy were less sanguine about the unannounced change of plans. Rather than wait for the appointment of a new advisory group. They requested a full report on the subject from Professor Bennett and advised him that he should not undertake such efforts. and intimated that a large part of it'could be devoted to this project. However. 'It was the con- sensus. A. Its chair. he had spoken lavishly of the resources that would be available to compute mathematical tables.9 The reality of the project proved to be quite different.This com- mittee had been founded in 1934.270 TABLE M A K I N G FOR THE RELIEF OF L A B O U R become anxious about Morrow. He had discovered a new sponsor for the project.

New York. After completing the degree in 1934. he had briefly tasted success when he was named one of the first post-doctoral students at the Institute for Advanced Study in Princeton.10 . Courtesy of Yeshiva University. Lyman Briggs found Lowan after he had left the institute and was shuttling between two jobs in New York. At night he augmented Ms salary by teaching physics courses at Brooklyn College. By day. Mathematical Table Makers. 1929. Upon arriving in the United States. engineer for a New Jersey utility. 10. After becoming a naturalized citizen in. he had worked as a combustion. Arnold Lowen (photograph circa 1938) was a professor at Yeshiva University.) neither of whom had ever done this type of calculation. 1948. (Picture taken from R. Archibald. He shunned the new and revolution- ary topics of quantum physics and turned to a more conventional subject.3 Arnold Lowan. New Jersey. He was an. Step by step he had mastered English and started to study physics. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 271 Hg. C. New York. where he had taken a degree in chemical engineering.The director of the project would be Arnold Lowan. he had enrolled in the PhD program at Columbia University. he was interested in calculation and willing to take the job when Briggs offered it. Jewish school on the north end of Manhattan. Though he had never overseen a large laboratory. immigrant from Romania. he was a professor at Yeshiva University. When asked to lead the programme. Scripta Mathematics. a. the cooling of the Earth.

he became increasingly curious how Blanch. . It gave her a enough free time to read mathematics on the side and so she registered for a course on relativity at Brooklyn College. After graduation. a former president of the American Mathematical Society. The work was pleas- ant and easy. handled correspondence and tracked orders. Her advisor had been Virgil Snyder.4 Gertrude Blanch (photograph mid-1930s) brought to the program a Ph. in mathematics from Cornell University and nearly twenty years of office experience. could do such fine work on the homework sets. One day he sat next to her on the bus and casually asked why she didn't attend class more regularly.D.) Lowan recruited Gertrude Blanch to be the mathematical leader of the project. she was an immigrant. and the late recipient of a doctorate. Blanch was a mathematics PhD who worked as an office manager for a photographical company. She had completed her degree in 1936. Devin Photographic. As the term progressed. 10. (Courtesy Deborah Stern." Lowan was the instructor. six months after her fortieth birthday. Blanch. who showed little inclination during the lectures she attended.272 TABLE MAKING FOR THE RELIEF OF LABOUR Rg. Like Lowan. the only job she could find was a one-year visiting position at Hunter College for women in New York City. she took an office job in a well appointed building on East 23rd Street. In the fall of 1937. When that job ended. She kept accounts. The firm. was flourishing during the depression through the unlikely strategy of selling a luxury good to the wealthy.

most of the computers were unable to do the work of the competent few.'13 However. which did prelim- inary computations for Blanch and the other mathematicians.12 Blanch had the task of preparing the work for the computers. She outlined how they would compute their functions. These indi- viduals were not merely out of work. Blanch did not let her com- puters work independently. The largest group of the computers would do only addition. First. she admitted that she held a PhD in mathematics. Only a handful would undertake the hardest operation. It was the west side of "West Side Story*. At that point. he was more direct. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 273 often a private individual. He took her to visit the offices for the new project.They were located in what was then Manhattans seedy west side. little about numerical analysis arid nothing about table making. She knew only the rules of commercial calculation. The most skilled computers were generally assigned to a special group. she concluded that this was her best opportunity to become a scientist again. of gang conflict. a neighbourhood just a few blocks from the Hudson River piers. as had much of the American mid- dle class. 'The New York high schools were pretty good in those days. The WPA offered few funds for supplies. Ever the optimist. he told Blanch about the WPA project and asked if she would be interested in being the chief mathematician. He complemented her work and asked about her scientific back- ground. After the next class. To less disciplined eyes. she restricted each computer to a single operation. '1 resigned from a beautiful office downtown' and took the position with the Mathematical Tables Project. she later recalled that some of the relief workers did fairly well. Blanch may have hesitated for a moment. A considerably smaller number would undertake multiplication and they would only multiply a single digit against a larger number. long division. and industrial buildings. Nonetheless. that it was something more than a shrinkage of the money supply or a loss of confidence in the stock market. The floor was covered with dust. . deflected his query. A few days later.' she com- mented. The offices for the Mathematical Tables Project were located on the tenth floor of an old structure. they could only be reached via a freight elevator. The furniture was broken and worn. how they would check the results and even how they would perform basic arithmetic.'We had a few very competent youngsters. tenements. the prospects were not promising. The next group would do subtraction. They had faced the full brunt of the depression and had learned the lesson. The story of her background intrigued Lowan. She dictated every step they took.

The curve had applications in statistics. they computed die first derivative f'-^(x) with the formula f(l\x) = —2xf(x).5). The number of derivatives varied at each point and was dependent upon the need to get 25 digits of precision. 10. eco- nomics. 3. First derivatives For each of these sixty values. For this step they used an old table of the H exponential function that had been published by the National Academy of Sciences.2) and a new calculation of F(0). Data which follow the bell curve are likely to take values near their mean (the central hump of the curve) and are less likely to fall at a distance {the sloping curves to the right and left that form the bell). Crude tabulation The computing floor began to tabulate the function F(x) using two formulae derived from a Taylor expansion of F(x): and Starting with F(0). they would produce pairs of values at increments of . they would use formula (1) to create F{. 2.l). and medicine. All calculations were done to 25 digits. 4. f(x) — —-?—J-J-'e ffd«i. Higher derivatives The special computers calculated higher derivatives using the formula /<">(*) = -2xf(tt~*](x) -2(n~2) f*"~~2}(x). psychology.1 .The second step would produce F(. which they would compare to the original F{0) to see if the calculations were . The bell curve describes the behaviour of certain kinds of statistical data. Initial vatucsThe special computing group calculated 60 key starting values of 2 the function. They always did at least seven derivatives and as many as twenty.274 T A B L E M A K I N G FOR THE RELIEF OF L A B O U R Computing the error function: steps the WPA Mathematical Tables Project used to compute the error function tables published in their Tables of Probability Functions Volume I published in 1941 The error function is derived from the famous beE curve (see Fig. electrical engineering. using both formula. a function given by the expression: The tabulation had six steps: 1.Then. The Mathematical Tables Project tabulated a version of the function which was derived from the right half of the bell curve.

A computer assigned to the addition group. In the course of a day. Final tabulation This step repeated the computations of step 4 with increas- inEply smaller increments. Her instructions for addition told the com- puters that they could add two red numbers or two black numbers but not a mixture of red and black numbers. xxi-xxiii.001 and finally at steps of . Once they were corrected. those that did were not allowed to use their own ways of implementing those rules. The third step would produce a F(. Columbia University Press. When faced with numbers of differ- ent colours. for example. Blanch prepared worksheets for the computers. would take a work- sheet. Mathematical Tables Project.3) and a new calculation of F(. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 275 accurate. they had to send the calculation to the subtraction group.01. She would print those values on the paper and would instruct the computers to compare . she required all computers to write negative numbers in red and positive numbers in black.l) to compare with the old values. the numbers were rounded to fifteen decimals and typed. New York. Tliey repeated the calculations first at steps of . then continued the process until they reached the fifth difference. Blanch and her staff" knew which values needed to appear in the final column of the worksheet. 'fiibtcs $f probability functions volume 1. then at steps of . The worksheets required the computers to match some of their values to pre-determined results. Source. For example. Subtabulation is similar to interpola- tion. 1941. The final typescript was subjected to a difference test using third and fourth differences. It starts with two known values and computes the numbers that fall between them. These sheets divided the calculation into its fundamental operations. Checking for errors The manuscript was checked with a fifth difference test. complete all the additions and then pass it to another group. Airy unusual values identi- fied potential errors.This procedure continued until they had computed F(6). For each computation. 5. Because of this. 6.OCX) 1. Many of the first computers did not fully understand the basic laws of arithmetic. Even. Blanch isolated each group of computers in a different part of the office and posted instructions for them to follow. Computers took differences of adjacent numbers (first differences) then differences of those differences (second differences). Most of the tables were computed through the method of subtabulation. pp. a sheet would pass between several groups as the calculations were done.

psychology. (Courtesy of the U. With applications in statistics. the computers would he responsible for correcting them.5 The error function is derived from the famous bell curve and describes the behaviour of certain kinds of statistical data. and medicine it was a common function and easily justified its inclusion in the Mathematical Tables Project Tables of probability functions. Errors could appear in intermediate values. To search for such . economics. Their worksheets were kept in large binders. 10. Mathematical Tables Project computers worked at long tables in an industrial building in New York. If the figures did not match. 10.S.) them with their results. Correct values in the final column did not guarantee the accuracy of the entire sheet.6 Computing floor of the Mathematical Tables Project (photograph circa 1940).276 TABLE MAKING FOR THE RELIEF OF LABOUR Fig. electrical engineering. National Records Administration. Fig.

'14 The claim that the tables were 'entirely free from error' proved to be over-confident. test was the difference test. even when all calculations have to be checked and rechecked multiple times. Though Blanch rarely chose the simplest methods. it examined the computing procedure itself to ensure that the human computers could actually do the work. Commentators joked that its acronym stood not for 'Work Projects Administration' but for 'We Poke Along. she tended to favour methods that minimized the number of calculations. who would solicit comments from mathematicians and scientists outside the project. It also examined the mathematics behind the computation and verified that it produced results as accurate as any other procedure. the mathematician John Curtiss. Certain patterns would indicate the presence of an error. Keeping the computers active was a crucial strategy for any project of the WPA. They were approved first by Lowan. and the pre- cautions taken seern to give considerable weight to the claim. Blanch and her assistants would test all the values on each work- sheet. the WPA had a reputation of supporting malingerers. Generally. Excessive . In reviewing the first two volumes of tables. They would check the work- sheet by taking differences of adjacent numbers on the sheet.*16 The job of keeping the com- puters busy was complicated by the procedure for approving projects. Nonetheless. Lowan didn't worry about the amount of effort the calculations required. They would derive a series of numerical tests in order to check for errors. A common. they also had to keep their workers busy. Rightly or wrongly. The computing plans were prepared by Gertrude Blanch and a small staff of mathematicians. The members of the Mathematical Tables Project were proud of their techniques for checking their work. Finally. the known errors in Mathematical Tables Project calculations is remarkably small. a procedure designed to give an unquestionable pedigree to the calculations. scientific institution. labourers who actively avoided work. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 277 mistakes.15 Selling to the scientists Not only did Lowan and Blanch have to present the Mathematical Tables Project as a professional. as errors did creep into the process from time to time. The approval process attempted to establish that the computations would be of general use and benefit more than a single individual or organization. A staff with more than a hundred computers is a surprisingly powerful computing organization. wrote "It is claimed that each volume is entirely free from error.

The books never mention the human computers. They shifted attention from how the tables were prepared to the mathematics behind them. written by outside mathematicians. described the computing plan. At this stage it was often difficult to find knowledgeable scientists who had not already offered their comments. the review process was really an extensive. More than once he wired the Washington WPA office reporting that the project was running out of work and urg- ing them to approve a computing plan. Lowan spent substantial effort in trying to recruit the most prominent scientists possible. DC for final approval. or engineers.18 The large tables. Lowan occasionally thought it so. physic- ists. The staff in the New York Office of the WPA commonly identified inconsistencies in plans and corrected mistakes. However. another round of math- ematicians would be asked to comment on the plan. Once the New York Office had approved the plan.17Yet. Briggs would send the plan to the New York Office of theWPA." Through their efforts. After receiving his staff's evaluation. These methods might require the subtle use of another table or a complicated re-use of intermediate values. the worksheets or Blanch's instructions. Lowan. and presented an analysis of underlying algorithms. highly ritualized version of peer review. They list only the members of the committee that prepared the computing plan. he would then submit the plan to Lyman Briggs at the National Bureau of Standards. Indeed. they would send it to the WPA headquarters in Washington. As before. These prefaces defined the functions to be tabled. The published reviews of the books uniformly praised the Mathematical Tables Project. After Lowan received comments. The outside reviewers regularly identified literature that had been overlooked by the project staff. She regularly rejected methods that were commonly used by skilled computers. Bureau scientists would give the project a second review and would try to assess the number of sci- entists who might make use of the calculations. Briggs and the staff of the WPA established a reputation for quality work. The WPA employed a member of the mathematics group at Bell Laboratories to review all scientific plans. which were published in distinctive tan colored vol- umes. this . included lengthy prefaces. Each of the reviewers commented on the plans and each added something to it. Briggs and his staff often had the best sense of the scientists that might use the work.278 TABLE M A K I N G FOR THE RELIEF OF L A B O U R calculations simply provided opportunities for the computers to make mis- takes. The approval process appears to be burdensome and ungainly. Blanch.

he was often promoting his own laboratory at the expense of the WPA.' In a charge that would resonate for the remaining life of the project. Punched card machines. although often in a lesser measure. In their mind. The most vocal critic of the project was Wallace J. 'In discussing a large project of table making one must consider whether the idea is to avoid work or to make it' he wrote with almost a mocking tone.22 Like Eckert. the project was a charitable organization. the human computers were relief workers. that he single-handedly saved the project at the start of the Second World War.'23 Lowan and Blanch were remarkably immune to the complaints of Comrie and Eckert. he wrote. Eckert would openly question the methods of the project. J. In a letter that must have surprised Lowan and Blanch. who was normally a strong supporter of the project. Lowan. This laboratory used punched card tabulators to compute mathe- matical functions. Comrie was a member of the MTAC committee. a laboratory barely two miles from the offices of the Mathematical Tables Project. In his private correspondence. he wrote that a proposed calculation seemed to be 'extravagant. He once attacked a proposed government laboratory by asking 'How much government can we stand?' He suggested that it would be best if the 'government undertakes nothing that can be done well other- wise. Eckert. not trained mathematicians. Eckert was the director of the T. Watson computing laboratory at Columbia University. to other work done by your project. His letters were . J. not a scientific laboratory.'are not well suited to the latter and hence are not recommended as a solution of the unem- ployment problem during a depression. A few scientists felt that the project was misguided. a former secretary of the British Association Mathematical Tables Committee (described by Mary Croarken in Chapter 9). For all of their success. and a former superintendent of the British Nautical Almanac Office. He was an enthu- siastic advocate for computation. His support for the Mathematical Tables Project was so strong that many of the project leaders came to believe. quite mistakenly.'20 Eckert was a strong conservative who liked little about the Roosevelt administration. Comrie. Comrie preferred computing machines to human computers. and to savour of com- puting gone amok. was unflappable. In framing his criticism in terms of the amount of work. he continued.'In fact this criticism applies. He also believed that some of the projects had little value and had been undertaken only as a means to keep the computers busy.'21 Such opinions were occasionally echoed by L. in particular. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 279 reputation could not stifle a lingering private criticism of the group.

testimonials which he carefully edited in order to present the Mathematical Tables Project in the best possible light. we must remember that he received little response for his efforts. Most of the work is of course done with the aid of calculating machines. 1 am much interested in your program and should like to get your material. Lowan's saviour was not von Neumann but Phil Morse. At times. he actually mis- led his potential collaborators. despite regular prodding from Lowan. the scientists didn't need to know about the social problems among the computing staff. He never mentioned epilepsy. Compton. or criminal records. "we are operating with a staff of 110 workers under the supervision of a planning section of which I am in charge.*25 At the time. a professor at the Massachusetts Institute of Technology. he wrote to a scientist.* he wrote. the project owned exactly three adding Monroe machines. It is easy to castigate Lowan for his unwillingness to admit that most of the calculations were done with red and black drafting pencils on graph paper and that few of his computers knew enough mathematics to balance a cheque book. has noted that scientists often feel that it is neces- sary to exaggerate the nature of their work in their attempt to gain the approval of the public.'27 Morse helped find work for the project. 'I may have some remarks and suggestions in connection with these things and will write to you concerning them before long. non-committal answers. Morse learned of the project in September 1938 through a circular that had been forwarded to him by a friend. John von Neumann. Occasionally. or the fact that many of the computers did not know how to subtract a bigger number from a smaller one. these letters present an unrealistically positive assessment of the group and ignore the real problems they faced. His contacts brought work that Lowan considered .24 Lowan clearly felt that he needed to do this in order to gain the attention of scientists. President of MIT that I should get in touch with you. who was just starting to become interested in computing sent a short note: 'Many thanks for the announcement of your project. Yet. 'I have been in charge of a number of similar projects here at Technology' he wrote to Lowan and he had been advised by 'Dr.'26 However. The his- torian George Daniels. The few scientists who replied usually gave cautious. In Lowan's rnind. von Neumann never offered any comments and never proposed any work for the project. During the first months of the project's oper- ation.280 TABLE M A K I N G FOR THE RELIEF OF L A B O U R filled with happy quotes from satisfied scientists. all of which were used by the planning committee.

28 In the end. All the military projects had to be approved not only by the National Bureau of Standards and the WPA but by the Army as well. which was credited only to Blanch and Marshak. It expanded the project to its largest size. the physicist and the direc- tor of the computing laboratory decided that they had no choice but to re-do the experiment so that it matched the calculation. several dozen used Sunstrand calculators. but also new equipment. a policy that apparently produced a con- frontation with Blanch. which Bethe adjusted. Once the work was over.31 . the first of these was a computation for Hans Bethe and his student. is usually cited as Bethe and Marshak but in fact. They required a detailed series of grid points to be computed for each map and they contracted the job to the Mathematical Tables Project. Blanch prevailed in the conflict and Lowan was not mentioned on the next paper to draw upon these computations. a second small group preparing tables. This expansion also brought a new layer of control. a punched card tabulator and a sorter. she completed the task in three weeks. In July 1941. It paid for three new Monroe calcu- lators. a group of 12 using Monroe calculators to do special projects. Lowan was reaching out to the military.30 While Blanch was working on the Bethe computations. The resulting paper. Robert E. the army was preparing a new series of maps. Blanch is listed as the lead author and Lowan is included as a co-author. Gertrude Blanch decided to do the calculation herself. another 100 using Sunstrand calculations. and a final group calculating grid points with the IBM tabulator. The contract brought not only new work to the project. A computing group at Indiana University had undertaken a large computation for a physicist ten years earlier and had used a constant later found to be incorrect. the group had 100 computers working by hand. Bethe was quite patient as each office looked at his plans. which is now regarded as a seminal contribution to the field. Marshak. In anticipation of a European war. Working with a few assist- ants. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 281 highly prestigious. The review found inconsistencies in certain constants. The request had to go through the usual approval process. Bethe requested a computation for a paper he and Marshak were preparing on the internal structure of the Sun. Such errors could be a major problem. rather than prepare it for the entire computing floor. In November 1938.29 Lowan insisted that his name appear on every publication from the project. assessed its use to a broader public and judged the correctness of his proposal.

they actually improved the stature of the project in the scientific community. many of the male computers left to join the military. A few of the most skilled computers found work doing computations for other governmental offices. This designation kept the project alive within the WPA but it did not guarantee the resources. Lowan succeeded in gaining the designation 'crit- ical war project' for the Mathematical Tables Project. which was preparing naviga- tion tables for the LORAN radio navigation system. Morse helped to arrange for the group to be divided between two organizations.33 For the duration of the war. From a single pair of stations. In practice. the group was saved by Philip Morse. the Mathematical Tables Project began to shrink. the Navy's Hydrographic office and the Applied Mathematics Panel. the mathematical branch of the Office for Scientific Research and Development. to keep the project at full strength. Their true position fell where the two curves intersected. The LORAN tables reduced the work to the simple process of looking up values in a book of tables. they would find a second curve. Computers left to take better paying war jobs in factories. It relied on pairs of radio stations to broadcast carefully timed signals. From a second pair of stations. the system required navigators to solve a complicated series of equations to find that intersection. human or financial.34 The preparation of the LORAN tables were one of the bottlenecks of the war. The greatest threat to the project came on 3 December 1942. After US entry into the war on 8 December 1941. The military found it difficult to deliver men and material until it .282 TABLE M A K I N G FOR THE RELIEF OF L A B O U R Shrinking for war As the country began mobilizing for war.32 Lowan again worked to find other sources of funding for his project but once more. The Army reclaimed their IBM equipment in July 1942. initially sixty-five in number. Lowan had no control over this group. in a separate office. when President Roosevelt announced that he was liquidating the WPA. and established them. After a flurry of letter writing and the inter- vention of Phil Morse. LOR AN was a joint project of MIT and Bell Laboratories. Navigators would record differences in these times and use them to identify their position on the globe. It required Lowan to reduce his staff and dispose of some of his equipment. Graphically the technique was simple. The Navy took the largest group of computers. navigators could determine that their position fell on a certain curve. As diffi- cult as these changes may have been.

He acquired calculating machines for each of his computers. Additionally. Abramowitz ran a race against time. and the distances on maps. Blanch prepared a textbook on advanced subjects of applied mathe- matics. the Applied Math Panel might have been forced to disband the group and assemble a computing organization with a more trusted back- ground. it circulated widely among the war offices involved in computation. and numerical analysis. for the members of his computing staff slowly left for better paying war jobs.35 The twenty-five were the best trained in the project and all worked with their own adding machine. after negotiations with the chair of the Applied Math Panel. she was asked to join the Manhattan Project in Los Alamos. The most elementary course dealt with the basic properties of arithmetic. Both Lowan and Blanch had applied for clearances but were denied. They computed trajectories of bombs.*37 Her words reflected a stoic way of viewing the circumstances. In practice. she and the sen- ior staff offered a series of six mathematics courses. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 283 had the navigation system in place. the LORAN group was producing a new table every three weeks with a staff less than half the size of the initial group. Were it not for the general nature of mathematics.36 The problem of security reminded the staff of the Mathematical Tables Project that they occupied a subservient role in the scientific infrastructure. The most advanced covered the theories of calculus. the leader of the LORAN computing group spent much of his time working to improve the efficiency of his organization. Milton Abramowitz. In 1944. which you didn't need. Published only in mimeographed form. the dis- persion of shock waves. The offer was withdrawn after her security review. he agreed to reduce his staff to a mere twenty-five computers. the transmission of heat. For this last course. As it's work came to a close in the spring of 1945. table making. it meant that the Mathematical Tables Project did the work of others but never initiated projects of their own. Lowan and the remaining Mathematical Tables Project computers became a general service organization for the divisions of the Office for Scientific Research and Development.3* . the lack of a clearance was more personal and frustrating.' argued Blanch 'The only thing that is classified is dimensions. No one in Lowan's group held security clearances. applied mathematics. reorganized the way in which the computers operated and identified a more efficient algorithm for preparing the tables. Blanch tried to increase the professionalism of the group by teaching them. Lowan had hoped to preserve a group of fifty computers for this work but.'In mathematics you didn't need clearances. During the lunch hour. For Blanch.

Almost immediately. Lowan returned to the rhetoric that had helped him in the 1930s. Freed from the constraints of the WPA and boasting a solid record of accomplishments from the war. To his surprise. felt that the group might be helpful in developing new computing machines and that there would be a dozen members 'who will be hard to replace once scattered. As the debate over the future of the project progressed. Lyman Briggs retired from the National Bureau of Standards.'40 That dozen included Gertrude Blanch and her assistants: Milton Abramowitz. OswaldVeblen. It re-unified the two parts of the Mathematical Tables Project and encouraged the computers to think that they had earned the respect of the scientific community and deserved to be a permanent organization. Jack Laderman. It would not include a majority of the staff and nor would it include Arnold Lowan. Ida Rhodes. The fate of Arnold Lowan summarized the struggle over the role of the Mathematical Tables Project. who was organizing post-war mathematical research. He argued vigorously for his ideas.39 Certain members of the Office for Scientific Research and Development agreed with that assessment. Lowan. and exaggerated the accomplishments of the group. However. In November 1945. located in Washington. Comrie.' a title that had come from their supporter and occasional critic. he paid the price for his aggres- sive efforts in promoting the project and for his inflated claims.J. It would also embrace a few of the computers who had completed Blanch's course of study. Lowan should have been able to find a good position for himself and his computers. DC. L. the mathematician . wanted to continue the group as a service organization. Edward Condon. lost his strongest champion when. They enjoyed thinking of themselves as the 'Mightiest Computing Team the world has ever known. Irene Stegun. he dis- covered that the post-war scientific leaders valued his project more than they valued him. a vision in which the Mathematical Tables Project would remain an independent scientific laboratory. invoked the aid of his allies.284 TABLE M A K I N G FOR THE RELIEF OF L A B O U R The mightiest computing organization in the world The war ended with a flush of glory and an anxiety over the future. Most were more interested in supporting the development of elec- tronic computers than in supporting table making. and a few others. Briggs's successor. he clung to a personal vision for the project. After the war. but their agreement had its limits. Herb Salzer. he indicated his plans by putting it under the authority of one of his assistants.

C. E. including Lowan himself. His vision of an independent com- puting laboratory succumbed to Curtiss' idea that computing should be subservient to groups doing scientific research. His wife and her family were from New York. the argument between Lowan and Curtiss was ultimately about the independence of computing organizations.41 Lowan immediately began lobbying Curtiss to keep his independence and to leave the Mathematical Tables Project in New York. His efforts con- tinued for nearly three years. The computers worked only a 32 hour week. WPA rules limited the hours of operation in order to let the workers search for a full time job in the private economy. He had written reviews of the first tables. Many of the workers. Witnesses to the conflict tended to cast the conflict in personal terms. as they had when they were relief workers. Curtiss had lost none of his respect for the project but he clearly conceived the group as a unit within a new national mathematics research group. It would serve the mathematical researchers and the scientists at the National Bureau of Standards. equally. where scientists could directly oversee it. a full-time position that he did not want to leave. by then. in corporate laboratories. Lowan made his case by trying to establish the professional nature of the Mathematical Tables Project. It would be a battle that would be repeated in the coming decades on university campuses. His appointment at Yeshiva University was. in the National Science Foundation. Curtiss had been a friend to pro- gramme. Like many of his claims about the group.42 Lowan. they generally saw scientific computing drift back to the laborator- ies. Ignoring the role of the WPA. exploited these rules to take a second job.Van Orstrand and John von Neumann. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 285 John Curtiss. the smallest of its four major divisions. becoming increasingly shrill. may have had some personal motivation for remaining in New York. he stated that it had been organized by the National Bureau of Standards and that he had been chosen to lead it by two well-known scientists. He first attempted to do so by arguing that the Mathematical Tables Project was the creation of existing scientific organizations. this statement was . angry. a struggle for control between Lowan and Curtiss.43 Even if it was motivated by personal reasons. Lowan lost this battle. may have concluded that he would have had limited con- trol if the group remained in Lowan's hands. In the early years of the war. The Mathematical Tables Project would be the Computation Laboratory of this group. Curtiss.. and accu- satory. Though many organizations would establish computing centres. and at the national research facilities. indeed.

286 TABLE M A K I N G FOR THE RELIEF OF L A B O U R based in truth and yet misrepresented the way in which the organization began operations. the Mathematical Tables Project did some calculations for the plutonium bomb. Their largest task had been to supply tables of Bessel functions to a Canadian metallurgist who was working on the problem of separating uranium isotopes. head of the Applied Math Panel. Indeed. Initially. Late in the war. but they were far from the centre of research. Morse was the more help- ful. Both Warren Weaver. In July 1947. though Lowan. The more powerful argument was to catalogue the quality of the pro- ject's work. including Phil Morse and John von Neumann. Again. the project had actually contributed computations to each of these groups but they had tended to work with individual researchers rather than the directors of these organ- izations. Lowen turned to his friends in the scientific community. the Fire Control Division of the Office for Scientific Research and Development. they were probably double- checking the work of computers at Los Alamos. The work had provoked a cer- tain amount of irritation between. the leaders of the new laboratory seemed cautiously interested in working with the project. the project provided tables of the binomial distribution. For the Army Signal Corps. As they began this work just three weeks before the first test of the bomb. weakened its impact by exaggerating his claims. the Army Signal Corps. which was then considered the epitome of war research. for example. and James Conant. He stated that the project had worked for the Navy Bureau of Ordnance. director of the National Research and Defence Council. He tried to find alternative sources of funding for the project. He enlisted von Neumann in his attempts to incorporate the Mathematical Tables Project in the laboratory organization. Furthermore. a common statistical model. had intervened in the corres- pondence to assure that the computations were done quickly and with lit- tle public attention. they had usually provided general mathematical tables which might be used in many kinds of scientific research. Lowan happily reported that 'The most fortunate solution to the problem of locating the Computational . as few had forgotten that the laboratory had begun as a work relief effort. Lowan and his client. The most promising seemed to be the new Brookhaven National Laboratory on Long Island.44 As before. the Mathematics Tables Project had made real contributions to the atomic bomb effort. Lowan soon abandoned this claim. Lowan tried to invoke the prestige of the Manhattan Project.

asked for a meeting with Curtiss and tried to bully him into changing his decision. which John is not. Morse confirmed that the New York office would shordy close: 'the situation you are up against is well nigh unbreakable.'His own words indicate that the proposed move is based on public relations considerations and the desire to make the Laboratory a Washington show- piece for Congressmen and other officials. C. while Curtiss would establish a laboratory in Washington. who had succeeded A. who is certainly a. gentleman. and fed them information about the move. John Curtiss informed Lowan that he would move the computing office in June of the following year.'47 The crisis lasted through early summer of 1948. the union that represented the human computers. Blanch left the pro- ject that spring to join the Institute for Numerical Analysis in Los Angeles. Lowan. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 287 Laboratory would be the scientific atmosphere of the Brookhaven National Laboratory. By early fall. Samuel Finkelstein. He also asked for support from the United Public Workers. advocated his removal on the grounds of'his brutal treatment of Lowan. Few believed that Lowan could maintain his base in New York. In August 1947. began looking for support outside of the scientific com- munity. growing increas- ingly desperate."48 In replying. the administrators at Brookhaven were raising concerns about the idea and soon thereafter the project stalled. It was a step that alienated him from other scientists.*46 As winter progressed into spring. Other senior staff members began packing their books and moving to Washington. The leader of the union. the rhetoric became heated and per- sonal.'The fact that you have not replied to my let- ter of November 5th. I don't like it but there it is. even Lowan himself conceded that the game was lost. He wrote to the faithful Morse. when the two sides reached a compromise. .'49 Lowan shut the door of the New York office for the final time on Thursday 30 June 1949. It was a compromise that offered little hope to Lowan. Perhaps the most vitriolic words came from R. Archibald. 1948 would seem to substantiate my feeling that you are seemingly unable to do anything which would change the trend of events. Bennett as the chair of the National Academy of Sciences MTAC committee. Lowan would have a year to find outside funding for a New York office. Archibald.'45 This hope was short lived. A. He contacted the New York Congressional delegation and asked them to pressure Curtiss. who had once a good working rela- tionship with Curtiss. By winter.

during social and organ- izational upheavals. the human computers were working on the first major trial of the simplex algorithm for linear programming. John von Neumann followed the trial closely and tried to estimate the time that the EN1AC would require for such a problem.288 T A B L E M A K I N G FOR THE RELIEF OF L A B O U R Epilogue The demise of the Mathematical Tables Project provides a striking example of how science can continue. At the height of the debate over the future of the pro- ject. The idea was revived at a 1952 confer- ence on table-making. but could not find the time or money to complete it. the book hides its lineage.51 The handbook is perhaps the single most widely circulated scientific work. but this is unlikely. as were a third of the contributing authors. a com- pendium of applied mathematics. The National Bureau of Standards estimates that about one million copies have been printed. a research program accepted by all in the scientific commun- ity. In all likelihood. In spite of its origins in the WPA. It did so with little time to spare. Though he had not initiated the calculations. twenty years after the liquida- tion of the WPA and fifteen years after the formal end of the Mathematical Tables Project. thus ending the last organizational trace of the Mathematical Tables Project of the WPA. the National Bureau of Standards may still have been sensitive to the idea that the group was nothing but an unqualified collection of relief workers. Most of the chapters are based directly on tables created by the Mathematical Tables Project. The book was published in 1964. The two editors of the book were veterans of the Mathematical Tables Project. the project had become what Lowan had wanted it to be. the Bureau of Standards incorporated the remnants of the project into its Center for Computer Sciences and Technology.The project out- lined several versions of such a book during the 1940s. One year after the publication of the Handbook.50 The Mathematical Tables Project finally reached a professional status with the publication of The handbook of mathematical functions. Conceivably. Further reading More detailed descriptions of the table making work of the Mathematical Tables Project and the National Bureau of Standards can be found in . at least temporarily. This trial represented a major change for the group as it was a problem that could not be tabulated and re-used later.

forthcoming. 'The process of professionalization in American science: The emergent period. J. Veblen Goes to War. 10 (Dec. For a general history of Roosevelt's employment policy dur- ing the depression. 922-931. Paul Brockett to Frank Lillie. 1. Scripta Mathematica 15 (1949).' American mathematical monthly 108. Studies in numerical analysis:papers in honour of Cornelius Lanczos. 29 October. 1974.'Table making at NBS'. Wlien computers wen human. Scientific Advisory Board Files. A. Office memorandum No. Rhodes. Archives of die National Bureau of Sciences. IEEE Annals of the History of Computing 20: 3 (1998). New York. Schlesinger. Houghton Mifflin. 433. p. Scientific Advisory Board Files. DC. p. The coming of the Neiv Deal. 151-66. 'Oswald Veblen and the capitalization of American mathematics'.).S. in B. Scaife (ed. Grier. Daniels. 1-10. 1820-1860'. 1937. Final Report of the WPA. tit. Academic Press. his 58 (1967). National Research Council Records. op. J. Archives of the National Bureau of Sciences. 474-97. 6. 1946. Princeton. ACM Press. D. National Bureau of Standards' in S. B. 607-21 and D. Grier. 1937. 33—63. pp. D.' A prehistory and early history of Computation in the U. Minutes of the Executive Council for the Physical Sciences Division of the National Research Council for 1934. 9. Archives of the National Academy of Sciences. Feffer. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 289 D. and G. Scientific Advisory Board Files. A history of scientific computing. Government Printing Office. Nash (ed. 7. Grier. Archives of the National Bureau of Sciences. Lowen. 5. Notes 1.The Math Tables Project: the reluctant start of the computing era'. 2. See also L. 'The mathematical sciences and World War IF. Office memorandum No. National Research Council Records. More wide reaching discussions of table making projects in the United States can be found in M. February 1996.'Gertrude Blanch of the Math Tables Project'. 3. Washington. A. . NJ.). G. Interview with author. New York. 433. Blanch and I. 8. see A. Princeton University Press. 'The Computational Laboratory at the National Bureau of Standards'. IEEE Annals of the History of Computing 19:4 (1997). 4. R. 237-50. 1959. Grier. National Research Council Records. Isis 89 (1999). 'Dr. American mathematical monthly 87:8 (1980). London. 39. Todd. 1990. 16 November. N. G. 2001). Rees. 33—50. Ross. E Lillie to E.

Stem Family Collection. The second volume. 33—50.A. the authors list (page 881) only ten errors in the Tables of the expo- nential function. A. For example. Commercial Research Department of IBM. Hayashi's 1926 book. Briggs.T. W. Interview with G. Archives Center. IEEE annals of the history of computing 20:3 (1998). Microfilm Reel 8659.und hyperbetfunktionen. has 60 known errors in 250 pages of tables. National Archives and Records Administration. Archibald. 19. Addison-Wesley. Eckert to D. 1962.0. D. Thatcher in San Diego.'Gertrude Blanch of the Math Tables Project'. W. In comparison.Tropp. et al. 20. conveyed by I. C. American mathematical monthly 48:1 (1941).'The Math Tables Project: The reluctant start of the computing era'. K. Record Group 69. Ida Rhodes to Ira Merzbach. 371—2. 20 July. At the far extreme. S. Lyman Briggs Correspondence for 1940. Charles Babbage Institute. 21.. Table of the First Ten Powers of the Integers 1 to 1000. University of Minnesota. Night Telegram from A. Papers of Phil Morse. 14. second edition. National Bureau of Standards Records on Electronic Computers. 11 January. 12. The approval process is best illustrated by a series of letters between Arnold Lowan. L.290 T A B L E M A K I N G FOR THE RELIEF OF L A B O U R 10. Milne-Thompson in June 1941. 1941. 17. 1946. H. 1 August. H. MIT Libraries. National Archives and Records Administration. Briggs. Rogers. Tropp. Administrative Records ofWPA Project 765-97-3-10. Papers of Wallace J Eckert. 16 September. New York City Office of the WPA. 1941. 1969. personal communication. 16 May. a book with 535 pages. Annals of the history of computing. 18. 15. An index of mathe- matical tables. Papers of Wallace J Eckert. Eckert to R. Gordy and L. 1973. The definitive list of errata is found in A. CB19. 1940-42. Morse papers. J. Lowan to L. Tables of higher mathematical Functions. Arnold Lowan to Phil Morse. Smithsonian National Museum of American History Computing Machinery Collection Supporting documents. 22. J. 16. is . Tables of die Exponential Function'. has about 2000 errors in 200 pages of tables. IEEE Annals of the history of computing 19:4 (1997). Rhodes. 13. Blanch by H. D. Smithsonian Interviews with Computer Scientists. 'Leslie John Comrie'.und mehrstellige taften tier kreis. H. 1-10. W. gives the most complete narrative of the project but it only describes the committee meeting on January 28. Reading Massachusetts. 1940. Box 1. Davis' 1933 book. Smithsonian National Museum of American History. Grier. Blanch and H. M. Sieben. Curtiss. A. 17 March 1989. 4 November. Subsequent volumes list the names of die senior mathematicians. 4:4 (1982). 56-7. Grier. This version of the story. R. Fletcher. Interview with G. J. CBI9. Special Collections. Charles Babbage Institute. Record Group 196. I. University of Minnesota. 11. Rubidge.Wallace. Tables of the exponential Junction.

32. Lowan to P. National Archives and Records Administration). September 1940. Administrative Records of the Mathematical Tables Project. A. Math Table Files. at its inception on 12 November 1942. 1948. Mathematics and Statistics. 3 April 1942. Lowan promoted Conirie's support as a reason to keep the panel (see A. Record Group 69. von Neunian Papers. Davis. Records of Office of Naval Research. GinnifF. G. 1962. Administrative Records of the Mathematical Tables Project. Records of the Federal Emergency Relief Agency (FERA). private printing. Morse. Project Number 765-97-3-10. 'The internal temperature density distribution of main sequence stars'. 1940-1942. San Antonio. von Neumann to A. However. 'The process of professionalization in American science. J. MIT Libraries. R. P. W Weaver seems to have kept his own council in regards to the Mathematical Tables Project (see W. his 58 (1967). General AMP Records 1942-46. 37-45. 273. A. L. 12 November 1942. Bethe. 26. A. National Archives and Records Administration. 8 February 1943. H. Marshak. 29 September 1938. 1946. long range navigation. Marshak. Daniels. at. E. G. N. Morse Papers. 1820-1860'. Records of the Federal Emergency Relief Agency (FERA). Grier. Morse Papers). T. 24. J.A. Weaver. pp. . Correspondence of the Applied Math Panel 1942—43. and H. Record Group 227). Blanch and R. E. 31.. National Archive and Records Administration. Record Group 69. 27. Project Number 765-97-3-10. Conant. D. T A B L E M A K I N G FOR THE R E L I E F OF L A B O U R 291 contradicted by the correspondence of the Applied Mathematics Panel. 19-20. Memo from Assistant Chief of Navy Bureau of Docks andYards. Morse 31 December 1942. WP. A. "Annual Report for 1941 "Administrative Records of the Maths Tables Project 1940-1942. p. New York. Project 765-97-3-10. et d. Journal of astrophysics. Box 17. Inception of the Applied Math Panel. 25. Comrie to J. Adventures of an ultra-aepidarian. 28. Furthermore. 151-66. Records of the Work Projects Administration for New York City. the members of the Applied Mathematics Panel discussed the importance of utilizing the Mathematical Tables Project (see report of W. Records of the Work Projects Administration for New York City. 1998 op. Pierce. 23. McGraw Hill. 29. L. Blanch. National Archives and Records Administration. Weaver to J. A. The emergent period. Lowan.Journal of astrophysics (1942). Box 46. 30. Records of Works Project Administration for New York City. 6 May 1942. 34. Lotan. Applied Maths Panel Records. Special Collections. J. Lowan. G. 33. Morse to A. Records of FERA. Lowan to P. 'The Internal Temperature Density Distribution of the Sun'. 21 September 1938. Record Group 69. 1940-1942. Library of Congress. Lowan.

Record Group 227. 44. Smithsonian Archives Center. A. MTAC Records. Lowan to P. Special Collections. Dantzig to J. AMP Records 1942—46. Morse.Tropp. 28 April 1948. similar language appears in a 11 May. Lowan to P. The book is uncopyrighted and it is difficult to give an exact total. 42. Gibbs. 39. Thatcher in San Diego. National Archieves and Records Administration. J. Blanch and H. there are editions published in foreign languages by foreign presses. In addition. Lowan to W. 18 July 2001. Applied Maths Panel Records. von Neumann Papers. 45. 1948 letter to Morse. 50. 26 July 1947. taught 1943—45 at the New York Mathematical Tables Project. MIT Libraries. 17 March 1989. Memo to Director. L. Special Collections. Physical Sciences Division. M. Morse Papers. Circular attached to S. Stern Family Collection. J.Archibald to R. Saber interviews with the author.Jordan. Special Collections. H. Hillman. 49. von Neuman. W. 13 June 1945. G. National Research Council Records. Interview with G. J. Special Collections. Annals of the history of computing 11:1 (1989). Todd. 5 January 1949. Correspondence 'D'. to A. 13-30. 'The Math Tables Project". Weaver. 9 April 1948. MIT Libraries. 'The National Applied Mathematics Laboratories—A prospectus'. 43. Lowan. 9 March 1944. AMP Records. Stern Family Collection. Quoted by I. mimeograph note attached to A. J. Executive Offices of Research Board for National Security. R. They believe that Dover publica- tions has sold 600000 to 900000 copies. 48. 47. Blanch. Morse Papers. Notes jor a class on numerical analysis. Veblen to L. communication with author. Oswald Veblen Papers. MIT Libraries. National Archives and Records Administration. Box 16. Morse Papers. MTAC Files. MIT Libraries.Archibald. FBI File of Gertrude Blanch. . Comrie.The Bureau of Standards has sold about 150000 copies. Morse. FBI. Morse.292 T A B L E M A K I N G FOR THE RELIEF OF L A B O U R 35. Rhodes in interview with H. 24 January 1949. interview with the author. C. Lowan. Todd. 41. Record Group 196. Rees to A. 37. ZG 227. P. G. Genera! Records. Curtiss. 46. 1945. 51. June 1998. 27 April 1948. Weaver to E. Blanch Papers. 15 February 1943. 40. Library of Congress. Morse Papers. July 19. National Bureau of Standards. 38. A. 20 September 1955. O. A. Correspondence of the Applied Math Panel 1. C. D. C. Finkelstein to R. 36.942-43. Moreland. Lozier. L. Archives of the National Academy of Sciences.

This page intentionally left blank .

18 SUM 1960 For 0h EPHEMERIS TIME .

1 7/?e Astronomical ephemerls for 1960. WILKINS Introduction A table of the computed values of the position of a moving astronomical body. (By kind permission of HM Nautical Almanac Office. 11 The making of astronomical tables in HfVi Nautical Almanac Office GEORGE A. Rg. The original title of The Nautical almanac and astronomical ephemeris was replaced in 1960 by the shorter title of The Astronomical ephemerls when the British and American almanacs for astronomy and navigation were unified and the work of computation and typesetting was shared. such as the Moon or a planet. which was first issued for the year 1767 to give astronomical data for navigation at sea. is usually called an 'ephemeris'. Nevil Maskelyne. The table on the feeing page was computed with the use of punched card machines and was typeset by hand.This word is also used in the title of The Nautical Almanac and Astronomical Ephemeris.) . At this time the Astronomer Royal. 11.

The computers and comparers were scattered around the country. and planets for each day of the year. or subtabulation as it is now known. Moon. which adopted the Greenwich meridian for the zero of longitude. was followed by other conferences that . When powerful electronic computers became available the use of precomputed tables and subtabulation were both replaced by direct evaluation of the trigonometric series of the theory of each body or by direct numerical integration of the equations of motion of a system of bodies. each coordinate is obtained by summing the contributions from the terms of a trigonometric series whose amplitudes and arguments are given by the theory.The coordinates were calculated from the precomputed tables at a wide interval and then interpolation by the method of finite dif- ferences was used to obtain the coordinates at the interval to be given in the ephemeris. The International Meridian Conference of 1884. every effort was made to reduce the number of multiplications required. The other principal technique that was used to reduce the amount of calculation was that of systematic interpolation. in particular. In the past each theory was represented by a set of tables that was designed to reduce the amount of calculation required to compute an ephemeris. This system was replaced in 1831 when the Nautical Almanac Office (NAO) was set up as a separate establishment in London. The techniques of computation used in the NAO and the design of the almanac changed only slowly during the nineteenth century. This technique continued in use even when loga- rithms were superseded by desk calculating machines and when commer- cial accounting and punched-card machines were adapted for scientific purposes. with a concentration in Cornwall. the process was very time-consuming. Consequently. This led to changes in the distribution of work amongst the staff and it allowed the expansion of the scope of the work to match the developments in dynamical and positional astronomy. These coor- dinates change slowly and regularly from day to day in accordance with the adopted theory of the motion of the body concerned. Flexible calculating machines and new formulae had to be developed before this process could be successfully mechanized. The principal tabulations in the Nautical almanac give the celestial coor- dinates of the Sun.296 TABLES I N HM N A U T I C A L A L M A N A C OFFICE organized the computations in such a way that each ephemeris was com- puted independently by two human computers and the results were then examined by a comparer whose task it was to resolve any discrepancies between the two. while Maskelyne was at the Royal Observatory at Greenwich. For an analytic theory.

to electronic computers and calculators entailed changes in the way that the work was organized and in the mathematical techniques and formulae that were used. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 297 led to international agreements in 1896 on the bases of the ephemerides and star positions and then in 1911 on arrangements for the sharing of the work of their computation.J. this article contains only a brief account of the techniques used for table making for the Nautical almanac from 1767 until 1925.3 Edwin Dunkin. Sadler. finally from 1959 onwards. Comrie developed new techniques so that National accounting machines and punched-card machines could be used effectively to produce tables for general use in mathematics as well for astronomical purposes. who was its Superintendent from 1930 to 1936. The printed page has been superseded for some applications by other media. give background . More new publications. however. there were changes in the extent and nature of the quantities that were published and in the forms in which the tables were made available to those who wished to use them. and publication-—are reviewed in this article. presentation. and new methods were introduced by the next Superintendent. Conine. states that he could find no record of the early computations. New techniques for printing and data distribution introduced during the second half of the twentieth century affected both the methods of production and the design of the tabulations. then to "difference engines' and. Not only did Comrie introduce new machines for use in the computa- tions.3 He does. The changes in the aids to computation from logarithms to desk calcul- ating machines. Consequently. but he also redesigned the Nautical almanac to make it more user-friendly and introduced new publications. H. while for other applications programs and data are pro- vided to enable the user to generate directly the numerical data that are required. Early procedures for preparing the Nautical Almanac The arrangements for the computation and printing of the early Nautical Almanacs have been described but as far as 1 am aware no examples of the original calculations for the Nautical almanac exist. Consequently. especially The Air almanac and auxiliary tables for use in navigation. such as the first in a series of volumes of Planetary co-ordinates. who was a Chief Assistant at the Royal Observatory at Greenwich.1 The pace of change in the NAO accelerated sharply after 1925 following the appointment of L. printing. These four aspects of table making-—computation. D.

These tables . (The overlap was due to the introduction of new fundamental ephemerides for the almanacs from 1984 onwards. A series of volumes of Planetary co-ordinates for the years 1900—1940. the NAO prepared a variety of sets of auxiliary tables for marine and air navigation. Confusingly. It was produced in cooperation with the Nautical Almanac Office in the United States Naval Observatory and was identical in content to The American ephemeris ami na. most of these gave the transformation from celestial coordinates to altitude and azimuth. The NAO has prepared The Star almanacfor land surveyors for the years 1951 onwards.ital almanac. From 1914 the navigational ephemerides were redesigned and published only in Tlie Nautical almanac. which then ceased to publish a separate almanac in German. The title The Nautical almanac has been used for the almanac for marine navigation since 1960.The title was changed to Tlie Astronomical ephemeris in I960.) In addition to the navigational almanacs. to 1981 both titles were changed to The Astronomical Almanac. From 1396 the tabulations for navigators were given in Part I of the almanac and this was published separately.298 T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E The publications of HM Nautical Almanac Office The Nautical almanac and astronomical ephetneris was published for the years 1767 to 1959. The scope was widened in the succeeding volumes of Planetary and lunar coordinates for 1980-1984 and 1984—2000. It also prepared annual volumes of Apparent places ofjundamental stars for the years 1941 to 1959. abridged for the use of seameni this title was changed to Tlie abridged nautical almanac in 1952. retained its original tide and the abbreviation (NA) until 1959 even though it was no longer intended for use by navigators. It was unified with The American air almanac from 1953. It was unified with the American almanac in 1958.The fall tabulations are still given in the American almanac. the main almanac. which gave high-precision data for astronomers. The Air almanac was first published for October—December 1937 and was continued until 1997. 1940-1960 and 1960-1980 were published for the primary purpose of tabu- lating rectangular coordinates of the planets with respect to the Sun for use in the calculation of the orbits of comets by amateur astronomers.ut. The responsibility for the publication was taken over for 1960 onwards by the Astronomisches Rechen- Institut in Heidelberg. From 1998 the daily pages were omitted from the British edition and its name was changed to Tlie UK air almanac. This was intended for general international use and it eliminated the need to include such data in each of the national almanacs that were then published.

and printing could be shared. who was appointed Superintendent of the Nautical Almanac in 1818. and trigono- metric and logarithmic tables. Dunkin describes the work of the Rev. Many different investigations were carried out and their results usually took the form of numerical tables.The almanac also included an ephemeris of the Sun so that the navigator could determine his local time and hence his longitude. The most important ephemeris in the early editions of the Nautical almanac was that giving the lunar distances (angles) of the Sun and bright stars from the limb (nearest edge) of the Moon. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 299 were also later unified with the American editions so that design. who was one of the computers for the early editions and who then became a comparer until his death in 1809. was less meticulous in his oversight of the work of the computers and comparers and so this task was taken over by Thomas Young. were away from their home ports for long periods. Sets of the NAO almanacs and occasional publications are held in the RGO archives in the Cambridge University Library. Seven-figure trigonometrical tables for every second of time (1939) and Five-figure tables of natural trigonometric junctions (1947). which were included in later editions of the Requisite tables. information about the computing arrangements that Maskelyne4 put in place by which two persons independently computed each ephemeris and a third compared them and resolved any discrepancies. The accuracy of the almanac was restored. SubtaMatiori (1958). The NAO also produced various mathematical tables for general use. espe- cially those on voyages of exploration. but many were written up in the form of NAO technical notes that had only a limited distribution. The almanac continued to give lunar distances up to the edition for 1906 even though accurate chronometers gradually carne into widespread use during the nineteenth century. in order to determine the Greenwich time of his observations. the navigator needed a set of Requisite tables. computa- tion. Malachy Hitchins. In addition. which had been written by Maskelyne. The most important were published formally. but Young ignored the requests for the inclusion of more data for use by astronomers. John Pond.They included Ititerpolatimt and allied tables (1936 and 1956). Maskelyne's successor. In particular. .3 The Nautical almanac was produced several years in advance as many ships.

Hind. Edwin Dunkin's father. Stratford was appointed Superintendent of the Nautical Almanac after the death of Young. but he received no support from Cowell. One member of the staff. who had assisted Hitchins from about 1804.6 The methods for the computation of the almanac remained largely unchanged for another century. a Leyton arithmometer. These make use of fourth-order interpolation formulae and include pre-calculated tables for the contributions of the differences. Hudson.8 Hudson is credited with the installation of a Burroughs adding machine in 1911. He also made a 'star-correction facilitator' to simplify1 the computation of the apparent places of stars and devised a double entry table for subtabulation using Everett's formula using a cork bathroom mat. R. however. who is best known for his dis- coveries of asteroids. M. Downing and P. and an early Brunsviga. S. As far as I am aware. keen to introduce new techniques. no examples of the computations during this period have survived. He was then the secretary of the Royal Astronomical Society which made many recommendations to the Admiralty about the content of the almanac in order to make it more useful for astronomical purposes. Some of the tables gave directly the logarithms of the trigonometrical functions in sexagesimal measure.7 Stratford was succeeded by J. Two members of the staff of the NAO who had known him wrote down their recollections of his work for me. was. Most of the recommendations of the Society were accepted by the Admiralty and were introduced in the almanac for 1834. H. who were transferred from the Royal Observatory where the computational methods were similar to those in the NAO. although some printed sheets give the formulae and precepts used for subtabulation. and then by A. C. Stratford changed the system of calculation by setting up the Nautical Almanac Office in London towards the end of 1831. Hudson published very little and so we do not know the full extent of his contributions to the making of mathematical tables. Printed multiplication tables and logarithm tables were used. moved from Truro to Camden Town in order to join the mainly new staff of the Office and he continued to work in the Office until his death in 1838.9 Comrie paid tribute to Hudson in a letter to .300 T A B L E S IN HM N A U T I C A L A L M A N A C OFFICE The computational work of the Nautical Almanac Office. Cowell was refused funding for a research group in celestial mechanics and from then on was content to allow the NAO to continue its long-established practices.W. Cowell.William Dunkin. 1831-1925 Lieutenant W. T.

Stratford 1831-1853 J. Comrie joined the staff* of the NAO in 1925 and became Deputy Superintendent in the following year.11 Twentieth century procedures for computation: electromechanical machines L.Wiikins 1970-1989 B. D.10 There are few records remaining of the methods used prior to 1926 as all past work and records were scrapped in March 1930.Comrie 1930-1936 D. and have been able to save many hours of work by the application of ideas that have been evolved from these inspirations*. Downing 1891-1910 P. He obtained more desk calculating machines in order to replace completely the use of loga- rithms and he made corresponding changes in the choice of formulae. W M. He wrote 'Ever since 1 joined the office I have felt most grateful to your husband for the successes which attended his efforts to get [the Burroughs] machine. S. Wallace 1999-preseiK Sinclair used the title 'Head'. He immediately proceeded to revo- lutionize the computing procedures of the Office and to redesign the Nautical almanac to make it more useful to astronomers. rather than 'Superintendent'. Hind 1853-1891 A. R. T. J. . Cowell 1910-1930 L. Mrs.J. and Wallace has continued this practice. A. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 301 The Superintendents of the Nautical Almanac Thomas Young 1818-1829 John Pond 1829-1831 W. The ideas which he applied to it were original and I have derived much inspiration from them.Sadler 1936-1971 G.Yallop 1989-1996 A.H. Sinclair 1996-1998 P. Hudson dated 1928 February 23.A. H.

13 His expertise became widely recognized and so he was able to obtain many dif- ferent types of machines for use in the N AO on trial. One of his first major achieve- ments was to use punched card machines for the summation of the hun- dreds of harmonic terms in the theory of the motion of the Moon. com- mercially as he was able to demonstrate the savings in cost. Each set-up specified the arrangement and functions of the 'stops' on the control bar. the sequence of the operations to be carried by the operator and the quantities to be printed. which were designed for commercial work. The Brunsviga remained the standard desk machine in the NAO until it was eventually replaced by electronic computers. These machines were used for checking by differencing and also to provide the differences that were required for subtabulation by the method of bridging differences. It was found that the manual Brunsviga calculating machines were better for general use than electromechanical machines. Each 'stop' deter- mined the way in which the contents of one or more registers were to be added to or subtracted from another register. rather than to attempt to design and build special machines for scientific purposes. In addition to their basic use for the summations in subtabulation.15 Comrie was also involved in the design. The Office acquired two machines. each with a capacity of 12 digits. Comrie held strongly to the view that it was best to make use of com- mercially available machines. two hours at a time. they were also used for summations in numerical integration. They had other uses and the collection of'National set-ups' contained the necessary oper- ating instructions for about 100 jobs. one of which used sexagesimal arithmetic. say. fourth differences. and summa- tion from.302 TABLES IN HM N A U T I C A L A L M A N A C O F F I C E He obtained the latest models of the Burroughs accounting machines12 and later he discovered the more flexible Ellis accounting machines. first seen at the Business Efficiency Exhibition in 1932. although the latter were faster for some jobs. computation. who would operate a machine for.14 The machines were operated on a rota system by scientific assistants. by obtaining the use of machines in the Admiralty or by hiring them. The Ellis machine. had four registers and could be used for differencing to. and publication of new mathematical tables with extremely high standards of accuracy and typography. He also introduced the use of punched card machines. The later National machines had 6 registers and a keyboard. which developed into the National accounting machines.16 The interval and interpolation aids were chosen to match the number of figures in the table so that the table would be easy to use. while .

and multi-register tabulator. reproducer. 11. The basic installation consisted of a sorter. The NAO continued to use National accounting machines but it also installed a powerful set of punched card machines. Comrie did not make a clear separation between official work for the NAO and private work for new general-purpose mathematical tables." After the war. and print the results. It could be linked to the reproducer in such a way that it was possible to punch the results on cards for input to the next stage of the computation. many of the new staff were transferred to the National Physical Laboratory to form the nucleus of the new Mathematics Division and in 1949 the NAO moved to the new home of the Royal Greenwich Observatory at Herstmonceux Castle in Sussex. was completed (apart from the card-controlled typewriter) in 1951 by the addition of an IBM 602A calculating punch. which was able to multiply and so could be used for tasks that up to that time had to be Fig.17 His young successor. Oxfordshire. D. The Office is now part of the Rutherford Appleton Laboratory at Chilton.18 During the Second World War the staff of the NAO was expanded so that it could act as the computing centre for the Admiralty Computing Service. The system. H. . so that he was suspended from duty in 1936. The Nautical Almanac Office was based at Herstmonceux Castle in Sussex from 1949 to 1990 as a department of the Royal Greenwich Observatory. TABLES IN HM N A U T I C A L A L M A N A C OFFICE 303 the typeface and the spacing of the figures and columns were chosen so as to minimize the risk of error in extracting data from the table. continued to expand the work of the NAO while maintaining the high standards set by Conine. collator. Each machine was controlled by a plugboard that was used to determine the ways in which the data were read from the input card and processed and in which the results were punched (on the same or another card) or printed. add and subtract them. which could read numbers from the cards into its registers. Unfortunately. Sadler.2 Herstmonceux Castle. with the 11 and 12 rows being used for minus and plus signs. These were of the Hollerith type with 80-column rectangular-hole cards.

This was within the capacity of our punched card machines. that were punched on the input cards. This was probably the last major use of the 'method of cyclic packs'. some small.c took about 2. One job with which 1 was associated—the evaluation of the series for the nutation of the Earth for a period of 100 years—took almost a year from inception to completion. It was also possible to use high-order differences for subtabulation with many figures. for monitoring progress. while the staff in the other sections were responsible for the overall planning of each job. some large. for providing the data. For example. which was based on the principle used by Brown in his Tables of the motion of the Moon and which was developed by Comrie for punched-card machines.304 TABLES IN HM N A U T I C A L A L M A N A C O F F I C E carried out manually. and for checking the results. This saved time.21 The use of these machines led to enormous increases in the productiv- ity of the staff.22 A major advance was made in the USA in 1953 when the IBM Selective Sequence Electronic Computer (SSEC) was used to carry out a long-term numerical integration of the coordinates of the outer planets. There was a gradual increase in the range of the tasks for which they were used.23 It was agreed internationally that this should be the basis of the ephemerides for 1960 onwards and the NAO had the task of computing the apparent geo- centric spherical coordinates from. the use of cards to hold intermediate results obviated the need for writing them down and then setting them for the next stage.5 seconds. but more importantly it reduced the number of errors that were made and so much less time was required to find and cor- rect errors. an operator might make an error in replacing a damaged card. so that fewer pivotal values had to be calculated. Scheduling of the work was a major task as many different jobs. In particular. 'Throw-back' methods developed by Comrie were used. the rectangular heliocentric coordinates. It had a very large plugboard with nearly 1500 sockets and it was possible to cany out a complex pro- gramme of operations in one run. would be in progress at any one time. as these allowed the effects of high-order differences to be taken partially into account by modifying the lower-order differences.20 The staff of the Machine Section of the NAO provided the technical expertise in the use of the machines. It was still necessary to apply independent checks to verify that the work had been planned and carried through correctly and it was still necessary to check by differencing that the successive values in an ephemeris were free from isolated errors. . It was very slow by current standards as an operation of the form a + b.

the machine showed clearly the positions where occultations would not be observable and so it obviated an enormous amount of exploratory calculation. My introduction to com- puters had been at Imperial College post-graduate lectures in 1950 and I later attended a short course on EDSAC at Cambridge in 1954. I had taken DEUCE programming man- uals to the USA. it was used for a wide variety of jobs . Apart from providing approximate predictions of the times of the occultations. but I returned to find that the Admiralty had decided that that the NAO should have a HEC 4 computer. Connecticut My main project was to make a new determination of the orbits of the satellites of Mars.2S I gained valuable experience in the use of IBM 650 computers in the USA in 1957—8. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 305 The Office also made use of an 'Decollation machine'. and for a short while the NAO provided a satellite prediction service for the UK. The machine remained in use even when electronic computers became available. Sputnik 1 was launched while I was in the USA. which may he regarded as an analogue computer that simulated the passage over the Earth of the shadow of the Moon as cast by a star. which was based on the ACE Pilot Model at the National Physical Laboratory. Nevertheless. which was renamed ICT 1201 by the time of its delivery in 1959. This computer. Twentieth century procedures for computation: electronic computers I was given the task of leading the programming effort when the NAO obtained its first electronic computer in 1959.The shadow was represented by a cylinder of light that moved over the surface of a rotating terrestrial globe on which the positions of a list of observers had been marked.24 The input settings represented the angular position and motion of the Moon with respect to the Earth. had so little capac- ity and speed that it did not allow scope for the development of new tech- niques for table making. that would otherwise have been necessary.26 Before I left for the USA we had reviewed the commercial computers that were becoming available in the UK and we had chosen an English Electric DEUCE computer. The operator turned a handle to move the shadow. rotate the Earth and move a needle around a circular dial that was read to give the times when the edge of the shadow passed over each of the observers. firstly at the US Naval Observatory and then at the Yale University Observatory in New Haven.

by staff throughout the Royal Greenwich Observatory. It was operated by NAO staff. . The ICT 1909 proved to be a powerful system but the computing demands of the Observatory rose rapidly. These computers were used by the NAO for work on the publications.3 The ICT 1201 computer was used in the Nautical Almanac Office from 1959 to 1964. The Computer Section of the NAO became the Computer Department of the Royal Greenwich Observatory in 1974 and the NAO became a minor user of the computing facilities after a VAX 117780 system was obtained in 1980 as a node of the national STARLINK network for image processing. The NAO did not. We found. The ICT 1909 was upgraded in 1974 to an ICL 1903T. 11. Again we did not get our computer of choice—we took delivery in 1966 of an ICT 1909 computer instead of an IBM 360 computer that would have given us compatibility with the computer at the US Naval Observatory.29 Nevertheless we still used dif- ferencing to verify that there were no random errors in the final results. for various occultation programmes. for computing the ephemeris of the Moon. while waiting for a decision on our bid for a better computer. we hired time on an IBM 7090 (later 7094) computer in London. it had a magnetic drum store of only 1024 words. but could not explain. California.306 T A B L E S I N HM N A U T I C A L A L M A N A C O F F I C E Fig.281 learnt how to program in Fortran by adapting a program writ- ten at the NASA-funded Jet Propulsion Laboratory in Pasadena. The machine is pictured here in the West Building of Herstmortceux Castle with operators Valerie Cann (on the left) and Lynn Ellis. have the resovirces to carry out major projects in the dynamics of the solar system. one isolated error. Instead the NAO concentrated on the improvement of the presen- tation of the tables and of the techniques for printing and publication. however.27 From 1963 onwards. and for some research in dynamical astronomy involving numerical integrations of the orbits of minor planets and satellites. A link to the ICL 1906A computer at the Atlas Computer Laboratory was installed in 1972.

A further factor was that the main almanacs were produced jointly with the Nautical Almanac Office of the US Naval Observatory. spacing. These factors are especially important in an almanac that is intended for practical use in situations in which speed and accuracy of use are both cru- cial. errors. (2) the choice of the interval and number of figures for each function. and interpolation aids——that are tabulated on each page or pair effacing pages. The factors that make a user-friendly table include: (1) the choice of the quantities'—functions. when its central processing unit was replaced by an ICL 1903T unit. Thus in an almanac for navigation an attempt is made to give in each opening all the quantities that vary from day to day. The mathematical tables produced by Comrie showed very clearly the benefits of using white space rather than rules to separate the columns and rows of the numbers. the demands of economics and the desirability of making all the required data available at one opening meant that this could not be applied throughout the almanacs produced by the NAO.4 The ICT 1909 computer was in use in the Nautical Almanac Office from 1966 to 1974. We found that the Americans were resistant to such . headings. Unfortunately. 11. while essential auxiliary data are given on end-flaps that can be seen at the same time as the daily data. (3) the overall design—layout. arguments. either for a particular day or for a group of days. Presentation of astronomical and mathematical tables There are several factors that determine the quality of a numerical table apart from the prime requirement that the values be free from. which in turn had to satisfy the US Navy. type size. typeface. TABLES IN HM N A U T I C A L A L M A N A C OFFICE 307 Fig. and (4) the availability of explanatory notes about the basis and use of the quantities in the table. and rules—of the page.

The differences in approach of the offices are clearly visible in the Astronomical ephemeris from 1960 onwards. however. sevens and nines are lower than the other numbers) in order to reduce the risk of errors in the reading from the tables. used throughout. it was reprinted many times. In general. On the other hand. Similarly. differences were not published if the table could be interpolated linearly. then the first differences would be printed. The mixture of styles was perpetuated in the Astronomical almanac from 1981 onwards even though it was printed only in the USA. but it did not discuss the procedures used for subtabulation. In general. fours. the user was expected to form the first difference mentally and then either to use a first-difference correction table or to calculate and apply the correction by using a desk machine. Several commercial publications were available.31 . such as those on the calendar and interpolation tables. Another important design factor that affected the ease of use of a table of a smoothly varying function was the choice of aids to interpolation. the almanac contained only brief descriptions of the quantities used by astronomers until 1931 when an extended 'Explanation' was introduced by Comrie. fives. as the first half was typeset in the UK while the second half was typeset in the USA. were published as separate booklets.308 TABLES IN HM N A U T I C A L A L M A N A C O F F I C E changes and so rules were often retained unnecessarily. if second differ- ences were significant. Comrie also split the numbers into small groups and changed from level figures (in which the top and bottom edges of the numbers are level) to head-and-tail figures (in which the tops of the sixes and eights are higher than those of the other numbers and the bottom edges of the threes. The scope of the auxiliary tables was expanded to include. Only rarely did the almanacs give functions that required the use of higher orders of differences. Some sections. It also gave formulae for 'differences in subdivided intervals'. for example. It was until recently the practice to publish separately the instructions and auxiliary data for calculating a time or position from the sextant observa- tions. This material was collected together in 1936 in the booklet Interpolation and allied tables. formulae and tables for computing derivatives from differences and for the numerical integration of differential equations. A completely revised and extended version of Interpolation and allied tables was published in 1956. publications that were designed and printed in the UK were printed without rules. while the Royal Navy had its own procedures.3" It included formulae and tables for the use of sixth differences as well as throw-back formulae. Level figures were.

The examples shown here are taken from the Sun's ephemeris for 1931. 1931. 1981 FOR O EPHEMERIS TIME Fig. 1973. (By kind permission of HM Nautical Almanac Office. The typographic style as well as the contents of the tables in the Nautical almanac and astronomical ephemeris changed over time. and 1981 and show changes in the typeface for the figures and in the extent of the rules used to separate headings and columns. 1960. 6 SUN.) . SUN. 1973 21 FOR O EPHEMERIS TIME C4 SUN. 11.5 Changing type faces in the Nautical almanac. 1960 19 FOR O EPHEMERIS TIME SUN.

Errors on the copy were still made even when it became possible to make up the copy by a scissors-and-paste . it included auxiliary tables and other reference data. This process. The first systematic account of the method was written much later by Wilkins for the booklet Subtabulation. were prolific sources of errors. It was published in the UK and was reprinted several times. In turn a truncated Chebyshev series could be replaced by an economized polynomial that could be evaluated by the use of a sim- ple recurrence relation without the need to look up interpolation coeffi- cients. which gives descriptions. proofreading. as well as information about many related matters. and the subsequent manual typesetting.34 At the time of the revision of Interpolation and allied tables there was a growing realisation that Chebyshev polynomials could be used to provide more strongly convergent interpolation formulae than those based on Gaussian polynomials at the expense of giving much higher errors outside the central region. The various sections of the book were drafted by members of both the UK and US NAOs.36 Copy preparation.32 It was developed further by Sadler and others in the NAO to meet particular circumstances. on 23 pages instead of 122 pages. and were then sub- jected to exhaustive criticism. This technique was first introduced in the Astronomical ephetneris for 1972 and it was used in the Astronomical almanac for 1981 in order to replace the lengthy Moon's hourly ephemeris. formulae and tables for several methods. checking and editing. The annual reprinting of lengthy sections of the Explanation of the Nautkal almanac was stopped during the war as an economy measure and it was not until 1961 that the long-promised Explanatory supplement35 was published.310 TABLES IN HM N A U T I C A L A L M A N A C OFFICE It appears that subtabulation by the method of'bridging differences* was first developed by Comrie.33 These procedures soon became obsolete as fast electronic com- puters made it feasible to compute the required values ab initio. given in decimals of a degree instead of in sexagesimal measure. It allowed the printing of equivalent data. Parts of it became obsolete before a completely revised edition was published in the USA in 1992. This gave the basis and method of calculation of the tabulations. and printing procedures Until the 1930s the printer's copy was prepared manually by copying the fig- ures from the computation sheets.

The editor of the publication was responsible for collating the errors that were found. the computed figures could themselves he in error. The code number on each set of proofs indicated the stage and type of reading. For many years the almanacs were printed from the moveable type—the printers used the Monotype system.38 The "main readers'. Examples were checked for appropriateness as well as for consistency with the formulae. The leading figures were examined separately as it was other- wise easy to miss an error. Users occasionally reported errors in the printed volume and these were listed in a later edition. One member of staff acted as the editor for each publication. The standards of proof- reading were monitored by noting when each reader missed an error that had been found by another reader. R a revised proof and S a stereo proof. Columns of figures that had been printed by a National machine or a punched card tabulator were cut out and stuck on pieces of card. Sometimes a further round of correction and verification was needed. The 'secondary readers' compared their proof with the copy by eye—not by listening to another person reading from the copy—and carried out such tasks as had been specified in the instructions for the page concerned. for returning to the printer one set of proofs on which the required corrections were marked and. mentally dif- ferenced the figures on the proofs in small groups looking for unexpected discontinuities. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 311 technique. and so the NAO used a combination of methods of proofreading.37 Headings and auxiliary data were usually inserted manually. later. who were usually the more senior staff. The main reader also verified that the tabulations were continuous from page to page and from year to year and he or she applied occasional inde- pendent checks that had been computed by someone other than the persons responsible for producing the original computations. There were several stages in the proofreading of each publication and at each stage at least four sets of proofs would be examined independently. typography. The printed differences were checked for con- sistency with the function values and were themselves differenced. so that P indicated a first proof. In addition the Superintendent would scan a set of proofs and would sometimes find errors that had not been found by either the main or secondary readers. for examining the next set of proofs to verify that the corrections had been made. rather than the Linotype system used . Moreover. The proofreading of text was also broken down so that spelling. The headings and minor items of auxiliary data were also examined separately. grammar. and technical meaning were verified separately.

The first such system was used at the US Naval Observatory in 1945 for the American air almanac for 1946. This had the disadvantage that new errors could be introduced during the printing processes after the proofs had been read. In addition the staff also came together for recreation. George Wilkins. new stereo moulds would be made. firstly. from the use of IBM card-controlled typewriters to provide copy that could be reproduced directly by photoli- thography. Lewes.312 T A B L E S I N HM N A U T I C A L A L M A N A C O F F I C E Fig. (By kind permission of the Sussex Express and County Herald. secondly. East Sussex. Mavis Gibson. Evelyn Grove. Preparing the Nautical almanac was a collaborative process. perhaps in places where no errors had previously been found. Considerable gains were made by using stereotnoulds. because of the complexity of the material. 11. or even after the corrections had been verified. A second round of main reading was carried out by a different reader on the stereo proofs in order to pick up any errors missed on the first proofs or any new errors introduced by the printer. from the use of better-quality printed output for printer's copy and.) for newspapers. which were made from the moveable type after it had been corrected.39 The first NAO system was installed in 1953 for use for the Abridged nautical almanac and the Apparent places of fundamental stars.6 A groyp of NAD staff after a tennis match 1953. Donald Sadler. and Aileen Grogan. Clockwise from top left are Gordon Taylor. Increases in productivity came. during the initial correction phase. If necessary. It was also used for some special publications such as .

The final step was the introduction of automatic composition by com- puter techniques. It was necessary to use preprinted forms that gave the page and column headings and the rules. Special procedures and great care by the operator were required to ensure that figures remained centred between the rules at the end of the page.43 The system proved to be so reliable that it was used for about 15 years before a new system was introduced. This system was replaced in 1971 by a UDS 6000 automatic typewriter. This system was used for the Star almanac as well as for the Astronomical ephemeris. It was sometimes necessary to paste on symbols. but none of them aimed to provide the sophisticated editorial control that was expected for the NAO almanacs. This may have been the first use of such a system in the UK.42 It was able to set headings and footnotes.We made no attempt to publicise our devel- opment of this system as we were too busy to write it up for publication. I built into the program all the rules that were then in use for manual composition. For our first run in 1968 we used paper tape to drive a Monophoto filmsetter for the Astronomical ephemeris I972. Phillips of HM Stationery Office. The computer output was on paper- tape and was used to drive a Monophoto filmsetter. usually. such as those for the phase of the Moon. The pages were photographed and reduced in size to. such as those for the suppression of common leading figures after the first line of a block of 5 lines. At the insistence of the Superintendent. A new system based on the IBM 370 document writing system was installed in 1963. I then developed a Fortran program that could be used more generally to set any required page layout. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 313 Nutation 1900—1959 And Planetary co-ordinates.44 . which was used for the editing of text as well as for tabular output.40 At that time other typewriters that were controlled by punched-tape did not give such high-quality output.41 The problems we encoun- tered confirmed our belief in the advantages of punched cards over punched paper tape. with paper-tape readers and punch. I believe that other systems were being developed in USA for printing tables. but was not suitable for text. He arranged for a student at the National Physical Laboratory to program in Algol the setting of one page from the Astronomical ephemeris. 70% for the production of the photolithographic plates. The first experiment was actually carried out in 1964 by Arthur H. we used a Linotton phototypesetter that was being used by HM Stationery Office to print telephone directo- ries for the Post Office and we were able to supply the data on 9-track magnetic tape. From then on.

ICSU) set the pattern for later cooperative activities. This was partly prompted by the space research programmes. star catalogues. for example. The importance of international cooperation in the production and distri- bution of both computed and experimental or observed data was recognized by other scientific unions and was organized by CODATA (ICSU Committee on Data for Science and Technology). In 1978. for example. there were major advances in the capacity of storage devices for data and in the soft- ware for the organization of databases and for the retrieval of information from them. International cooperation in this matter was required since it was necessary. Astronomers shared in the development of these techniques. For example. which needed existing astronomical data and which also supplied new data. in an attempt to reduce costs.314 TABLES IN HM N A U T I C A L A L M A N A C O F F I C E Publication and distribution media As far as I am aware the printed page was the only medium for the distribu- tion of the ephemerides and other tables of astronomical data until after the Second World War. the IAU appointed me as chairman of its Working Group on Numerical Data and its delegate to CODATA. The Jet Propulsion Laboratory. . became the major generator of high-precision orbits of the Moon and planets. By I962. The next step was the use of 7-track magnetic tape and then 9-track magnetic tape for the interchange of star catalogues and astronomical ephemerides. In 1970. During the following nine years while 1 held those positions.45 Each 15cmX 10 cm microfiche held 98 pages. the NAO had built up a library of ephemerides. The increasing level of cooperation between the almanac offices led to exchange of almanac data with the US Naval Observatory and later with other offices. the data were also made available in machine-readable form for those who wished to analyse the data in depth. and observational data on punched cards and supplied copies on request. to agree on stand- ards for the interchange of such data. microfiches were used to publish a catalogue of observations of occultations of stars by the Moon. the heliocentric coordinates of the outer planets that were computed in 1953 in the USA by numerical integration were supplied on cards so that we could compute the geocentric coordinates to be published in the almanacs. Nevertheless. The first examples of the exchange of data on punched cards were probably for star catalogues. At its first meeting in Prague in 1969 a joint working group of the International Astronomical Union (1AU) and COSPAR (Committee on Space Research of the International Council of Scientific Unions.

T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 315 The increasing use of small programmable electronic calculators led to the production of a new type of publication that allowed the user to bypass the ordinary printed almanac for the data needed for astronavigation. Further reading A biography of Maskelyne along with an authoritative account of the early years of the Nautical almanac is given by Derek Howse. which gives data and both Mac and PC software on CD-ROM for longer periods for most of the ephemerides in the Astronomical almanac itself. but the users were expected to key the data into their hand-held programmable calcula- tors. but the technique was clearly preferred to the manual use of the more expensive main almanac by many navigators. It was prepared by HM Nautical Almanac Office and the dedication reads: This book is dedicated to the Royal Greenwich Observatory (RGO) which closed during the preparation of this fifth edition. Cambridge University Press. The development of small computers. the user may obtain directly the required data for any particular time or place without the need for interpolation. which is a convenient.46 but later volumes were published by HMSO.. floppy discs. This was both a chore and a source of error. the current issue of Compact data con- tains a CD-ROM with the NavPac program for astronavigation and much more information than the printed text and tables in the booklet itself.49 Moreover. It is poignant to note that the RGO was originally established in 1675 in order to '. Although the user now has the means to pre- pare his own almanac it nevertheless seems likely that there will still be a role for the old-fashioned printed almanac. 1989. led to the use of cassette tapes. The first booklet giving data for 5 years in compact form was published by the RGO in 1981 for trial purposes. and later of the ubiquitous PC. For example.. and now CD-ROMs for the distri- bution of astronomical and astronavigational data and of software that can be used with the data.47This booklet made use of economized polynomials. Howse also gives .48 A. such as the BBC microcomputer (which was tried in the NAO in 1985). reliable source of much of the information that is required. find out the so-much-desired longi- tude of places for the perfecting of the art of navigation'. further development by the American NAO is represented by the Multiyear interactive computer almanac (MICA). Nevil Maskelyne: the seaman's astronomer.

A survey by P. (1976) 243—67. Oxford University Press. dis- tribution and storage of data led to new problems for astronomers who had previously relied on large printed volumes for their catalogues of numerical data on stars and other astronomical objects and phenomena.S. Dick (ed.316 TABLES IN HM N A U T I C A L A L M A N A C OFFICE explanations of how lunar distances were used for the determination of lon- gitude in Greenwich time and the discovery of longitude. published by Philip Wilson. of the basic formulae of celestial mechanics and of the procedures for the numerical integration of the equations of motion precedes a historical review of the use of computers (especially in the US Naval Observatory and related organizations) for celestial mechanics and for typesetting is given in J. Wilkins. This volume contains many papers on the history of the US Nautical Almanac Office and the methods employed there. Jaschek & G. Wilkins (eds. Marcel Dekker Inc. Naval Observatory. An overview of the history of the Nautical Almanac Office is given by G. critical evaluation and . 239-43. Kent. Washington. 4. illustrations. 'The expanding role of H M Nautical Almanac Office. with many additional. DC. G. Holzman and A. who was Director of the US Nautical Almanac Office. A. London. A. Vistas in Astronomy. Encyclopaedia of computer science ami technology. 1980. Although primarily concerned with observational data. 1999. (New edition 1997. Fiala & Steven J.). 1999. 1968. March 3-4. Wilkins has also given a general account of the history and work of the Office in a paper with many references and a chronological list of the main events between 1. The most relevant are C.). HMSO. vol.) An introductory account of the development of the method of lunar distances and of later methods for position finding at sea and in the air is given in Man is not lost: a record of two hundred yean of astronomical namgation with the Nautical Almanac 1767-1967.). pages 54—81. Kenneth Seidelmann. US Naval Observatory.767 and 1998 in Alan D. London. New York and Basel. Compilation. Oxford University Press. the databases that were then being developed. A. The availability of computers and new methods for the acquisition. Belzer. while M. 20. provides an overview of Comrie's work at the Nautical Almanac Office and else- where. Proceedings: Nautical almanac office sesquicentetmial symposium: U. many of the new techniques were relevant to the storage and distribution of ephemerides and the results of computations on models of stars and stellar systems. Croarken's Early scientific computing in Britain. The proceed- ings of the early conferences to discuss these problems highlight the import- ance of (a) measures to ensure the quality of the data and (b) techniques to retrieve data from. 1818-1975'. 1976. 1990. (eds.

. Personal communications by W. Cambridge University Library RGO 16. 1977 and C. 10. Hingley & T. A Cornishman at Greenwich Observatory. Croarken. together with an explanation by D. Automated data retrieval in astronomy. See Chapter 1 of P. an introduction by the editors and 6 appendices.Jowrrw/ of the Institute of Navigation (London). See. 1999. 9 (1886). 4. G. Royal Greenwich Observatory Archives. This spiral-bound book (218 pp. Cambridge University Library RGO 4/34. Notes 1. 'The bicentenary of the Nautical Almanac'. 5. E. HMSO. Sadler. No. or the books listed for fur- ther reading. Truro. 7-18. H. 2. 21 (1968) 6—18. 64. 391-401. Proceedings of IAU Colloquium No.) contains a transcription of the auto- biographical notes of Edwin Dunkin (pp. An account of international cooperation in the production and publication of astronomical ephemerides is given in the introductory chapter of the Explanatory supplement to the astronomical ephemeris and the American ephemeris and nautical almanac. 6. Dunkin. Smith on 6 April 1966. 'Notes on some points connected with the early history of the "Nautical Almanac" \Joumal of the Royal Institution of Cornwall. forthcoming.). Sadler. 'The foundation and early development of the Nautical Almanac'. p. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 317 distribution of stellar data. Royal Institution of Cornwall. The papers of Joshua Moore (one of Maskelyne's computers) in the Manuscript Room of the Library of Congress in Washington contain some examples of the prepared noon and midnight lunar distances for die computer to interpolate between but do not show details of the workings. 18:4 (1965). Dordrecht- Holland: Reidel Publishing Company. box 54. July—Sept. 8. Forbes. 35. 3. Cambridge University Library RGO 16/54. 9. London. C. Jaschek & W. E. A. A far off vision.). Dordrecht-Holland: Reidel Publishing Company. 'Tabulating the heavens: Computing the nautical almanac in eighteenth century England. for information about die first comput- ers of the Nautical Almanac. There is an example of Hudson's 'cork mat' in the Royal Greenwich Observatory Archives. See Maskelyne's notebook in Royal Greenwich Observatory Archives. Proceedings of IAU Colloquium. Journal of the Institute of Navigation. a foreword by Allan Chapman. for example. 1961. Scott on 21 February 1966 and by E. 1981. 7. 29—172). H. 25:3 (2003). Heintz (eds. IEEE Annals oj the history of computing. Daniel (eds. See also M. D. D.

A. NAO. 92 (1932). Comrie. 16—18 for notes on L. This booklet includes a photograph and a description of the occultation machine and its use. Mathematical table makers. The prediction and reduction of occupations. 15. in Improved lunar ephemeris 1951-1959. Cambridge University Library RGO 16. London. Yale University Press. The application of the National accounting machines to the solution of first-order differential equations'. 'Coordinates of the five outer planets. 32 (1991). E. Wilkins. New York.'On the construction of tables by interpolation*. 25. L. 433-41. H. Comrie: a forgotten figure in the history of numerical com- putation'. J. 1937.'Calculation of the nutation from the new series'. Monthly notices of the Royal Astronomical Society. Comrie. Smith. Sadler. New Haven. Comrie. The correspondence relating to Comrie's suspension from duty is held in Royal Greenwich Observatory Archives. 11. Sadler. 87-114. 1948. J. 88 (1928). 506-23. W Brown. 16. Quarterly Journal of Applied Mathematics 1:4 (1948). 1919. G. 523-41.J. 'The Nautical Almanac Office Burroughs machine'. Porter. box 40. Archibald. Journal of the British Astronomical Association. Todd and D. Carter & D. E. with a list of his principal tables.'Punched card machines at Herstmonceux'. Royal Greenwich Observatory Archives. Cambridge University Library RGO 16/60. D.J. HMSO. Brouwer. 420-2. 114-18. 2:19 (1947). 1954. 14. 185-9. 19. XII. 13. 22.'Admiralty Computing Service". O. 20. 17. Supplement to the nautical almanac for 1938. Cambridge University Library. H. . Naval Observatory. Royal Greenwich Observatory Archives.B. J. The application of the Hollerith tabulating machine to Brown's tables of the Moon'. Eckert. Croarken. Scripta Mathematica. Washington.J. L. L.'Donald Harry Sadler. Comrie. A. Tables of the motion of the Moon. G. See pp. J. 21.S. 36:4 (April 2000).E. C. Astronomical papers prepared for the use of the American ephemeris and nautical almanac. Journal of the Royal Statistical Society (Supplement) 3:2 (1936). Mathematical tables and other aids to computation. Comrie. 1653-2060*. vol. Monthly Notices of the Royal Astronomical Society.318 T A B L E S IN HM N A U T I C A L ALMANAC OFFICE 10. 61:7 (1951). box 63. Monthly notices of the Royal Astronomical Society. Mathematics Today. Quarterly journal of the Royal Astronomical Society. US Government Printing Office. 694-707. Royal Greenwich Observatory Archives. (1908-1987)'. U. Personal communication by E. 24. M. R. 12. M.'Inverse interpolation and scientific applications of the National accounting machine'. Cambridge University Library RGO 16/15. A. 18. L. and G. Clemence. DC. 289-97. W. 67-8. G. 23.'L. 92 (1932). Wilkins.J. RGO 16. 1953. J.

Series A. 'The prediction service of HM Nautical Almanac Office*. boxes 40 and 411. box 35. Explanatory supplement to the astronomical ephemeris and the American ephemeris and nautical almanac. 185—95. HMSO. 31.'Subtabulation with special reference to a high-speed computer'. RGO 16. NAO. 30. 32. London. P. Cambridge University Library. H.). Quarterly journal of mechanics and applied mathematics. The introductory section includes bibliographic notes on methods of subtabulation. Royal Greenwich Observatory Archives. 1961. Cambridge University Library. The HEC 4 was aimed at the commercial market and had no scientific software. RGO 16. Comrie. K. NAO. HMSO. Sadler) describes 'The method of precalculated second differences'. Printed from the stereo plates for the Nautical almanac for 1937. 45-8. Cambridge University Library. T A B L E S IN HM N A U T I C A L A L M A N A C O F F I C E 319 26. 87-114. 1992. London. Part III (by D. London. University Science Books. 27. It merged with another company to form International Computers and Tabulators (ICT).Woollett. Sadler. 1956. 28.'Inverse interpolation and scientific applications of the National accounting machine'. while Part IV (by G. California. Mill Valley. L. Part II of this booklet gives tables for some 'Direct methods'. HMSO. Interpolation and allied tables. boxes 61 and 222 (file 13R). A generalized technique for the use of bridging differences was given by Enid R. 29. 1958. 36. Guidance and informa- tion about its use was given by Wilkins and later by others in a series of lec- tures and in NAO computer circulars. RGO 16/52 for collections of proofreading notes. Interpolation and allied tables. The white card was cut from obsolete Admiralty charts while the glue was sold under the name of'Cow Gum'! It was desirable to open the windows while it was being used. Supplement to the journal of the Royal Statistical Society. D. Royal Greenwich Observatory Archives. See Royal Greenwich Observatory Archives. Some years later it merged with English Electric to form International Computers Limited (ICL). Explanatory supplement to the astronomical almanac. 33. Proceedings of the Royal Society of London.J. This was a com- pletely rewritten version of the 1936 edition. Its main store was a magnetic drum with a capacity of 1024 words. 248 (1958). 34. See Royal Greenwich Observatory Archives. NAO. The HEC 4 computer was manufactured by the British Tabulating Machine Company (BTMC). A.Wilkins) describes 'The method of bridging differences'. Seidelmann (ed. . which had supplied most of the punched-caid machines in the NAO system. London. 3 (1936). H. 35. It was considerably extended in its scope. Subtabulation: a companion booklet to interpolation and allied tables. Cambridge University Library RGO 16. 37. 1936. 38. HMSO. II (1958).

Hohenkerk and B. Nautical Almanac Office in: Royal Greemvich Observatory: Telescopes. D. D. RGO 16. 183. . 16 pages and 5 microfiches. 132 pages and a CD-ROM. 46.Y.. 85. Richmond. Royal Greenwich Observatory Archives. Morrison. B. US Naval Observatory. box 8. research and services: October 1 1980-September 30 1985. instruments. HMSO. E Haupt. Eckert and R. 1998. Hohenkerk.The Stationery Office. RGO 16.1981. J.'Compact data for navigation and astronomy for the years 1981 to 1985*. Cambridge University Library. 44. Cambridge University Archives RGO 16. pages 92—3. 41.V. Royal Greenwich Observatory Archives. Inc. Cambridge University Library.Yallop. Royal Greenwich Observatory Archives. Cambridge University Press. Compact data for navigation and astronomy for the years 1986 to 1990. Willmann-Bell. undated. 40. NavPac and compact data 2001-2005. 2:17 (1947). boxes 22. 48. 197-202. files 40P and 101/8 (IV) contain further notes and correspondence about the development and use of the system. 47. B. L. RGO. Mathematical tables and other aids to computation. RGO 16. the first Astronomer Royal. Astronomical Applications Department. Virginia.22. C. 43. 'The printing of mathematical tables'. 45. Multiyear interactive computer almanac 1990—2005 (MICA).Yallop. This report also includes paragraphs on the changes in the Astronomical ephenteris for 1984 and on new publications and techniques.320 T A B L E S IN HM NAUTICAL ALMANAC OFFICE 39. Royal Greenwich Observatory bulletins No 185. 1985.Y. 49. Royal Greenwich Observatory bulletins No. The new system is briefly described in the report on the work of HM. W. 2000. D. London. 42. London. boxes 5. Roya! Greenwich Observatory Archives.1978. Yallop and C.'Catalogue of observations of occupations of stars by the Moon 1943-1971'. The quotation is from the royal warrant for the appointment by Charles I! of John Flamsteed.

This page intentionally left blank .

you've seen. ORDER OF RECALCULATION So far.V1SICALC® 3S? IBM PERSONAL COMPUTER LESSON FOUR TUTORIAL The screen should look like the following photograph: OQS'OZWf" If you would like to use this version of the budget worksheet for your own use. As a rule. but you haven't been told much about how this is done. Some characteristics of recalculation can affect results on a complex worksheet. An entry at position Al cannot be a formula that references other positions. Each formula is evaluated only once unless you ask for an extra recalculation by typing !. The VisiCalc program recalculates by starting at the upper left corner of the worksheet. working its way down and to the right until it reaches the lower right corner. that the VisiCalc program recalculates the values of all the formulas on the worksheet. save it by typing /SSbudget «J. . this means that formulas that reference other entries must be located below and to the right of the referenced entries.

Logarithmic tables have now been made obsolete by the electronic calculator. (Reproduced with permission of International Business Machines. The decline in the use of tables was much in evidence before the arrival of the electronic spreadsheet on the scene. as a paper-based artefact. Examples of these uses of tables have been given in several of the pre- vious chapters.) . It explained an unfamiliar technology and provided user assistance at a time when computer memories were too small to support on-line help. Tables had two main uses—as a calculating aid (such as a logarithm table) and as a data storage device (such as an actuarial or census table). and by and Fig. 124 User's guide lor VisiCalc. 1981. Is close to the end of its technological life. while data tables are increasingly being replaced by online data- bases. 12 The rise and rise of the spreadsheet MARTIN CAMPBELL-KELLY The mathematical table. c. The User's guide was a necessary complement to the early spreadsheet program.

Similarly designers of civic structures now understand that a horizontal surface at about knee height affords sitting—which they may or may not want to encourage. depending on which way the door opens. while a smooth vertical surface encourages graffiti—which they probably do not want to encourage. one can argue that they sprang from similar historical and technological forces. and these have not evolved to any significant extent. Seen in this light. Even before the invention of paper. . Hence the spreadsheet can be seen to make use of two-dimensional affbrdance in exactly the same way as a table. it is clear from Eleanor Robson's chapter on Mesopotamian tables that the clay tablet lent itself to two-dimensional expression. Conversely. but has two additional properties-—easy erasure. and the ability to act as a 'window' onto a much larger virtual surface. the perception of the spreadsheet as an historical successor to the table is intu- itive and appealing. Aflbrdance can be defined as: A bundle of properties about some specific object that provides us with an oppor- tunity to perceive something specific or to move through the world in a charac- teristic way. a writing surface affords the property of organizing information in a two-dimensional grid. So.1 Thus well-trained architects now design doors with either handles that afford pulling or flat plates that afford pushing. but to have the additional properties of dynamic alteration. a two-dimensional writing surface does not afford easy three-dimensional representations of data. and quasi-infinite extent in the plane.324 THE R I S E AND RISE OF THE S P R E A D S H E E T large the spreadsheet has not taken over these functions of the table. It seems plausible that the two-dimensional table would arise spontaneously in any civilization where a writing surface was used. The screen of a personal-computer shares the two-dimensional character of a writing surface. Yet. while one cannot assert that the mathematical table is the direct historical ancestor of the spreadsheet. In what sense can this said to be true? Spreadsheets vis-a-vis tables I believe one answer to the above question is offered by the psychological concept of'aflbrdance' which has been much used by architects and human computer interaction designers in recent years. and therefore tables can be viewed almost as an historical inevitability.

such as ease of use. all worksheets should be preserved. by asking the same kind of questions that one might ask of tables. its use avoids delays in presenting these financial statements to management. A worksheet was (and is) a standard tool for constructing a trial balance. work sheet). In a similar way. whether made by human computers or printers. be completed before the adjusting and closing entries are made in the books. Even though the worksheet is not part of the formal accounts. How were spreadsheets produced and sold? Who were the users of spreadsheets? Within these larger themes. How have spreadsheet designers married form and function to assist the user? Babbage. THE R I S E AND R I S E OF THE S P R E A D S H E E T 325 Because tables and spreadsheets share these fundamental properties. which was originally called 'an electronic worksheet. in this chapter the table-centric view of Sumer to spreadsheets will be used as a lens through which to examine the spreadsheet. and the problem of plagiarism. errors.2 How have spreadsheet makers sought to deter would-be copyists? Prehistory The Library of Congress catalogue lists over a thousand texts on spread- sheets published during the last 20 years. as a trap for plagiarists. According to a textbook first published in 1956: The information needed for a formal balance sheet and income statement can be obtained quickly by means of a work sheet. was obsessed with eliminat- ing the possibility of errors occurring in tables. Comrie was famous for the fact that he introduced small rounding errors in his edition of Chambers six-figure mathematical tables. and how this could facilitate their use and minimize errors. an astonishing testament to the extent to which they have entered the commercial and scientific culture. Accountancy textbooks published up to the 1970s typically have an index entry for worksheet (or. above all. although the former is much more prevalent. and many of his successors. . rather. As the work sheet can. Babbage.' Both of the terms worksheet and spreadsheet have their origins in accounting. and to minimize the potential for error. As Mary Croarken has noted elsewhere. spreadsheet makers have gone to considerable lengths to make their products easy to use. were fascinated by the visual presentation of tables. but not for spreadsheet. Very few of these books suggest that the spreadsheet has a history. There is no record of who introduced the word spreadsheet into the computer lexicon. but it was not the inventor of VisiCalc. some secondary issues will be explored.

In the 1960s and 1970s many software packages for financial analysis were designed for use on mainframe computers and commercial time- sharing systems. could be purchased from stationery suppliers. And just as the human computer had large sheets of paper feint-ruled into quarter-inch squares.3 Hence the worksheet was both an aid to systematic calculation and an auditing device. it now appears that the computer has progressed to the point where it provides management with useful tools for the planning process. Two of the most powerful tools available are finan- cial modeling and "what if" budgeting. These early tools originated in the discipline of 1960s management science. so an accountant was to a spreadsheet. spreadsheet had acquired a meaning that can best be explained by a simile: as a [human] computer was to a tabulation. This package was a dialect of the BASIC programming language for financial analysis: One of the most useful applications of BBL is for report generation: preparing and formatting reports. not only in audits by public accounting firms but also in those made by govern- ment tax examiners in connection with income tax returns. The technique became so standardized that pre-printed worksheets. Thus a writer of 1972 noted: After years of merely processing clerical transactions. strategic planning.5 There were three main problems with these financial analysis packages.6 Second. One example of such a system was the Basic Business Language (BBL) offered by the popular Tynishare system.4 The term 'what-if analysis'. sometimes called spread sheets.326 THE R I S E AND RISE OF THE SPREADSHEET These are generally required by outside auditors in any examination of the books. By the 1970s. the programs had poor interactivity so that results would at best be available after a minute or two using a teletype terminal attached to a time-sharing . Such packages were primarily used by middle-management for financial forecasting. also appears to date from. typically in 6 or 10 column format. the early 197Os. the accountant had spreadsheet paper ruled into a grid for recording financial data. are often used to summarize historic information or forecast future performance. running a program on a time-sharing system would have cost at least $10 an hour. For example the median price in 1974 of a main- frame financial analysis package was $2800. which itself has a deep historical connection with post-war operational research and long- range planning. and budgeting. the term. much used in the spreadsheet literature. alternatively. they were expensive. Reports of this type. First.

Sales were only modest at first. reach- ing 12000 copies a month by the end of the year. and usually required more than a modicum of computer programming knowledge. In short. the price was increased to $249.7 Bricklin was a 26 year-old Harvard MBA student who had at one time been a software designer with the Digital Equipment Corporation. 12. and the program he and his partner Bob Frankston produced was sold asVisiCalc from late 1979. a level the market was evidently willing to bear. and the idea for a Visible calculator* was born of that experience. Bricklin. The detailed history of how the spreadsheet was invented and kick-started the personal computer revolution is a standard item in every account of the rise of the personal computer. to write the pro- gram.2 Dan Bricklin inventor of the personal-computer spreadsheet demonstrating VisiCalc at the West Coast Computer faire.) . Third.VisiCalc went on sale at a retail price of $99 in October 1979. where he worked on. the packages were difficult to use. Bricklin and Frankston incorporated a development company Software Arts. or at worst after several hours when using a conventional mainframe. The acknowledged inventor of the electronic worksheet was Dan. Cambridge. Fig. Bricklin. He was exposed to financial analysis techniques using tedious pencil and paper methods. 5 December 1979. (Reproduced with permission of Bob Frankston. At the same time. enrolled for his MBA as a springboard to a career in the financial services industry. Mass. and used a software publisher Personal Software to manufacture and dis- tribute it.. but following favourable press reviews and word~of~ mouth recommendations sales took off during the second half of 1980. word processing software and newspaper typesetting systems. among other things. THE RISE AND RISE OF THE SPREADSHEET 327 computer.

8 VisiCalc's ease of use arose from the fact that the user did not need any pro- gramming knowledge. Although personal com- puters were still relatively expensive in 1979. interactivity. the machine was able to keep up with the thinking speed of the user. The real-time interactive nature of VisiCalc was not just an improvement on main- frame financial planning software. but in VisiCalc the user simply entered values into the worksheet and specified the math- ematical relationships between them.. VisiCalc's 'instantly calculating' nature was a revelation to seasoned computer users. where the user directly arranges the print on the page. none had exactly the properties of VisiCalc. Jerrold Kaplan. as opposed to a typesetting system where a set of instructions is used to lay out the text. It was able to recalculate quickly the rows and columns of a spreadsheet every time a single number was changed.9 While it took hours or days to learn even rudimentary computer pro- gramming.. It turns out that some problems are easier to solve using the procedural approach. in traditional programming terms. but they were not real-time. the Chief Technology Officer of the Lotus Development Corporation put this very well: While some spreadsheet users may not recognize it immediately. Indeed. you write the program as a description to be maintained (a set of relationships between data elements). . and unprecedented ease of use. Rather than write a program as a procedure to be executed (a series of well-defined steps to be performed in a particular sequence). . known as the procedural and declarative program- ming paradigms. Mainframe financial modelling packages were often on-line systems. such as processing orders or updating a database.328 THE R I S E AND RISE OF THE S P R E A D S H E E T Whether or notVisiCalc merely provided features already available on existing financial analysis packages was of less importance than its cheap- ness. Because the response time was virtually instantaneous. In this respect VisiCalc was analogous to a word processor. while others are easier to solve using the declarative approach. this process is actually a method of programming. it was a new paradigm. Bricklin sometimes described VisiCalc as a word processor for numbers. such as projecting profit and loss or performing a lease-versus-buy analysis. they were very cheap compared with corporate mainframes and mini-computers. anyone with a basic mathematical knowledge could use a spreadsheet straight out of the box. a point well articulated by an analyst of 1985: Although there have been financial modeling and planning programs on main- frames. In financial analysis packages such as BBL the user had to code the problem. The difference between these two problem solving styles is a recurring theme in computer science. VisiCalc took care of the rest.

IV.. the user can see only a 'window' on the spreadsheet of perhaps 8 or 10 cells wide by 20 deep.The number of rows was as few as 256 in early spreadsheet programs. and TOTAL represent column headings. THE R I S E AND R I S E OF THE S P R E A D S H E E T 329 Anatomy of a spreadsheet A B C D E F 1 Reading Writing TOTAL 2 Janet 22 21 43 3 James 14 18 32 4 John 19 17 36 5 AVERAGE 18. percentages. X. A value cell contains quantitative information that can be used in a computation.. while Reading. some labels are centred while others are aligned left or right. while others arc given with two places after the decimal point..B4 is shown as 18. while the table as a whole occupies the rectangular range of cells A1.B. separated by a double period. Cells can contain one of three entities: a label. Values are usually decimal numbers.The rows are numbered 1. James.C4 are all integer values entered by the user. Thus the AVERAGE of the range of cells B2. . while John's TOTAL is in cell D4. A range of cells is denoted by die start and end cells of die range.3. AZ. BB.. the six central cells B2.. Thus in the fig- ure the labels Janet. but they can also include currency values. AA. John.. Individual cells are denoted by the concatenation of the column and row indices: thus in the figure the name Janet is in cell A2. and AVERAGE are row labels.AC. In the figure..AB.. which cannot enter into a calcula- tion and is usually used to name neighbouring cells. of which the top left corner is shown above.33 18. some values are shown as integers. Most spreadsheets have 256 columns indexed by a one or two letter combination:A.. or A formula: A label consists of a textual item. Writing. The contents of a cell can be formatted appropriately by the user: in the example above... Z.D5.2.. At any time. tip to many thousands on a modem' program^ It is difficult to conceive of applications that would be able to make use of such a large worksheet. or even dates. A worksheet consists of a matrix of cells.. .33 in cell B5. BA.. C.Y.6? 6 The basic form of the personal-computer spreadsheet has remained essentially unchanged since it was first introduced inVisiCalc in 1979. a value.

in cell C4. just one spread- sheet tended to dominate with 70 per cent or more of sales. Microsoft Excel. each of which had to be supplied with a spreadsheet. allowing for very complex analysis. The single most important feature of the spreadsheet program was its ability to immediately show the effect of changes.VisiCalc had sold 100 000 copies by May 1981. The production and diffusion of spreadsheets Spreadsheets are a huge business. being the first program of its kind for the then market-leading Apple II computer. For example. Some of these were quite successful. early-start advantage. and several other early computers. was changed from 17 to 19. 20 million copies a year. the average score for Reading is evaluated in cell B5 by the formula @AVERAGE(B2. VisiCalc had an. and an estimated 700000 copies by late 1984. .B4). It is possible to use pre-defined arithmetic and commercial functions (sometimes known as @functions in early spreadsheet programs because of the syntactic @-sign that preceded their name). A formula cell defines a computation involving other cells. when it had been on the market 5 years. which were largely incompatible. subse- quent spreadsheets used similar (though not always identical) notations. The diffusion of spreadsheets was dramatic from the very beginning. In die figure above. Hence within two or three years of the launch of VisiCalc there were dozens of what the trade press termed 'Other Gales'. For example. VisiCalc was subsequently re-written for the Radio Shack TR. Much of die analytical power of modern spreadsheets is achieved by incorporating dozens of functions for statistical calculations and financial analysis. not the formula itself.330 THE R I S E AND RISE OF THE SPREADSHEET Anatomy of a spreadsheet cont. and only a handful of products have ever achieved real prominence. Thus in the figure. if John's score for Writing. cell D2 evaluates the sum of Janets Reading and Writing scores. Note that die spreadsheet displays the result of the formula. cells D4 and C5 would instantly change to 38 and 19.S-80. was selling more than.. During the early 1980s there were many competing personal computers.10 By the mid-1990s the leading spreadsheet. 'Visi-Clones'. InVisiCalc this formula would have been expressed as -4-B2+C2. or 'Calcalikes'.11 The market for spreadsheets in highly concentrated.33 respectively. A formula can use the results of other formula cells. such as Sorcim's SuperCalc and Microsoft's Multiplan. however. On any single platform.

As with the earlier platforms.. IBM announced its personal computer. Because the marginal cost of software production is trivial compared with the original development cost. bundling lacklustre products with market leaders. incremental sales generate very high profits. and in creating application. Network effects refers to the fact that a successful package creates a community of users who share . and so on. These profits can then be reinvested in product improvement. The theory runs along the following lines.Then the market will not stabilize at a 50:50 split as one might have expected. led by the economist Brian Arthur. the magic of Microsoft economics kicks in. Thus VisiCalc dominated from 1981 to 1983. These 'switching costs'. advertising. files. and lower prices. besides the cost of the new package itself. user lock-in. Tactics include free trials. say). The leading protagonist of the theory of increasing returns is the Stanford School. THE R I S E AND R I S E OF THE S P R E A D S H E E T 331 In August 1981. 50 split is unstable and the market will tend to tilt slightly in favour of one product or the other (spreadsheet A. Once market dominance has been gained. Both of these make it very difficult for users to switch from their current spreadsheet program to a competitive one. and Microsoft Excel from 1990 to the present. but one product will eventually dominate strongly. The theory of increasing returns-—sometimes called'Microsoft Economies' in business magazines—explains the tendency for a single product to domi- nate many high-tech markets. this makes product A. and network effects. Lotus 1-2-3 from 1984 to 1989. subsequently known simply as the 'PC*. Economists have a number of explanations for the existence of these 'serial monopolies': the theory of increasing returns. competitive upgrades.The reason is that a 50. Today soft- ware marketers have become very adept at 'tipping' an unstable market in their favour when introducing a new product. a single spreadsheet dominated the IBM-compatible PC market. Once product A has achieved a small sales advantage. include the lock-in factor of the intellectual investments a user has made in learning to use the program. but especially to software in which increasing-returns behaviour is most pronounced. Suppose there are two equally priced and equally desirable information products (say spreadsheet A and spreadsheet B). this advantage is sustained by user lock-in and network effects. This created a standard personal computer platform for the first time. more desirable and get still further ahead of B. although this changed over time.12 The theory applies to many high-tech products. and soon IBM 'clones' accounted for 80 per cent of new computer sales.

Thus 1-2-3 users could present their results in compelling graphical forms such as bar and pie charts. formerly known as Personal Software) was close to bankruptcy. and they could use information permanently stored in a database.332 THE R I S E AND RISE OF THE SPREADSHEET files. In the case of the IBM-compatible PC there have been two such discontinuities. Users simply selected what they perceived to be the best spreadsheet on the mar- ket. when the Microsoft Windows operating system achieved market suc- cess and was supplied with virtually every new PC. Lotus 1-2-3's ascendancy over VisiCalc was absolute. For a couple of years Lotus was perhaps America's fastest growing major company. Hence. users wanting to take advantage of the new operating sys- tem had to buy Microsoft Excel—the only spreadsheet for Windows then . Because the costs of switching from one product to another are so high. and they became employees of the firm. and by 1987 it was the world's largest inde- pendent software vendor with over 2000 employees and revenues of $400 million a year. users usually only make a change when there is a 'platform discontinuity*. or no computer at all. almost entirely derived from sales of 1 -2-3. For example.VisiCalc was the first and dominant spreadsheet for it. In January 1983. Because Lotus 1-2-3 would not run under Windows. Lotus 1-2-3 dominated IBM-compatible PC spreadsheet sales until 1990. and their ability to co-operate depends on them using the same package. while Bricklin and Frankston's development com- pany Software Arts was taken over by Lotus Development. it is even more difficult for a group of workers to switch products than an individual. the Lotus Development Corporation introduced its 1-2-3 spread- sheet.* Despite the fact that 1-2-3 cost $495 compared with $249 for VisiCalc. Needless to say manufacturers of the dominant software packages have become accomp- lished at making the switching costs as high as possible. In August 1981. The arrival ofWindows was another platform discontinuity. it massively outsold it. or at least the same file format. when people were beginning to buy the IBM PC en masse. In addition 1-2-3 ran three times as fast as VisiCalc and was more 'user friendly. it included graphics and database capabilities in addition to the spreadsheet function (hence the name 1-2-3). Lotus 1-2-3 was a major improvement overVisiCalc. By 1985 Visi- Corp (the publisher of VisiCalc. people came to the IBM PC from many different non-compatible computer models. and so there was no lock-in effect on their purchasing decisions. However. when the IBM PC was introduced. while competitors work hard to make the change as painless and cheap as possible.

introduced in January 1983. direct manipulation by means of a mouse. and powerful aids for charting. graphical data analysis. integrated three functions—a spreadsheet. The figure shows the data from the spreadsheet in the previous text box presented as a bar-chart. 12. The new program made a leap forward in terms of ease-of-use and data analysis.3 Lotus 1-2-3.4 Microsoft introduced Excel 2. Spreadsheet evolution: Lotus 1-2-3 and Microsoft Excel Fig. and a simple database. Fig.0 for the IBM-compatible PC in autumn 1987. The figure shows the data from the spreadsheet in the previous text box presented as a bar-chart. it provided the user with multiple windows. 12. .

to exit VisiCalc one was required to type the sequence of characters VSQY'. By 1990. When it was introduced in 1979. the increased code size had been accompanied by vast improvements in data analysis and presentation. such asVisiPlot andVisiTrend that produced visual representations of spreadsheet data for incorporation in written reports and presentations. Alongside improved data manipulation and presentation. The tight integration of all these functions within a single program was very attractive to users. and ease of use. One index of the advancement is the size of the programs themselves. over the years spreadsheet publishers have made massive investments in product improvements. However it quickly became apparent that users were keen to do more with their spreadsheets andVisiCorp intro- duced a number of complementary products. early versions of VisiCalc look prim- itive in the extreme. Release 3 in 1989 enabled users to manipulate several spreadsheets simultaneously. Microsoft Excel con- tained about 650000 lines of code.13 Although frequently criticised as 'bloatware'. VisiCalc consisted of roughly 10000 program instructions. it was not until the platform discontinuity of Windows that it was able to capture significant market share. Lotus 1-2-3 made an . Lotus 1-2-3 was systematically enhanced throughout the 1980s. release 2 in 1985 introduced sorting and enabled much larger spreadsheets to be manipulated. However. Compared with today's spreadsheets.334 THE R I S E AND RISE OF THE SPREADSHEET available. ease of use has been a major selling point of spreadsheets. Lotus 1-2-3 eliminated the need for such complementary products by incorporating the same facilities in the basic software package. For example. Product improvement and perfection As increasing-returns theorists would predict. This change was made as painless as possible by Microsoft offering price discounts for Lotus 1-2-3 users and by the product supporting Lotus- compatible file formats and facilities. It is worth noting that Microsoft's spreadsheet Multiplan had been available since 1980 and in improved form as Excel since 1987. VisiCalc's user interface now seems archaic. For example. VisiCalc originally manipulated and displayed nothing other than text and numbers on a monochrome screen. Users were provided with a 'Pocket Reference' card with many such command strings.

Users find themselves buried beneath an overwhelming weight of wizards. three frus- trated authors were even driven to write a book on the topic. By the mid-1980s there were less than 20 significant products. At the time of writing.15 It seems that every feature. ethnographic field studies. THE R I S E AND R I S E OF THE S P R E A D S H E E T 335 advance in user friendliness by introducing menus in English. far beyond the needs of any one individual. partly to keep up with competitors. Excel 97 annoyances. Lotus 1-2-3 and Excel. one simply had to select a menu and chose the 'Quit' command.1* And yet. as Cusumano and Selby have shown in Microsoft secrets. usability laboratories. telephone help-desks. There is no subject on which users are so vocal and so at odds with the manufacturers as the phenom- enon of 'creeping featurism*. Nonetheless there is no ques- tion that each product release adds new capabilities to the software. the num- ber of spreadsheets was remarkably small. and feedback from. are all in about their tenth major release. With the introduction of Microsoft Windows and Excel it became possible to use a mouse to perform many operations by direct manipulation. One might say that owning a spreadsheet and some . However. and the only other sig- nificant spreadsheet Quattro Pro. This is perhaps the inevitable outcome of designing one-size-fits-all software for a community of a hundred million users. There is a commonly held suspicion that the introduction of new product versions at approximately 18-month intervals is as much about revenue generation as genuine product improvement. Complementary products Compared with the population of different mathematical tables. or provided 'templates* for particular applications. and today there are just three. problem solvers. and far beyond what any person who is not a full-time developer could master. Indeed. their excellent academic study of Microsoft's development processes. to quit a program. These software packages enhanced or adapted the spreadsheet by extending its capability. enhancements to products such as Excel are largely driven by user research-—focus groups. is indispensable to a significant group of users. in the 1980s there developed a secondary market in complementary programs for the best-selling spreadsheets. Now. and partly to exploit new technological opportunities such as the World Wide Web. and 'macro* facilities that go far. however apparently obscure.

) . scientist. While every engineer. each profession had its own set of specialist tables. (Reproduced with permission of Funk Software. eliminating the need for scissors and glue. or actuary owned the same standard logarithm tables.336 THE R I S E AND RISE OF THE SPREADSHEET complementary products was rather like owning a book of common logar- ithm tables together with a few volumes of special purpose tables. 12. Sideways enabled a spreadsheet to be printed across several pages of tractor-feed stationary.5 One of the most popular complementary products for Lotus 1-2-3. Fig.

Noteworthy. However. seeing the add-on market as parasitic and something it would like to keep in-house. By 1987. A package called The Analyst enabled errors in a spreadsheet to be detected. third party suppliers quickly began to make 'add ons* to enhance 1-2-3 in new and unforeseen directions. Sideways was a real boon for users and it sold hundreds of thousands of copies.A. In 1985 Lotus began to actively co-operate with third-party producers by supplying an 'Add-in Development Kit*. this was another best seller. instead of across. Another. mentioned previously. a product known as Sideways produced by Funk Software enabled a spreadsheet to be printed along the length of continuous stationery. Although Lotus 1-2-3 on its first release in 1983 incorporated all that VisiCalc and its complementary products could do within a single package. Every one of these products had several competitors. Indeed it was possible to have several add-ins attached to 1-2-3. At first Lotus did not actively co-operate with complementary product suppliers. Second. By contrast an add-in was a program that could be tightly linked with 1-2-3. D.V. and soon there were scores of add-ins from dozens of suppliers. First it increased user lock-in. it soon became apparent that the add-on market had two major benefits. but also a user's collection of add-on products. specialized add-ons enabled 1-2-3 to satisfy many different user communities in a way that no mass- appeal product from a large firm. For example. While an add-on program complemented 1-2-3. could. a phenomenon that Forbes maga- zine call the 1-2-3 'after market'. for now a competing spreadsheet would have to supplant not only 1-2-3.E. so that the two behaved as a single integrated program.16 For example. enabled elec- tronic Post-It notes to be attached to the cells of a spreadsheet. one product SeeMore enabled the spreadsheet to be reduced in size so that more of it could be seen on the screen. . In 1988 Funk Software introduced a new product Allways that added desktop publishing capabilities to 1-2-3 to exploit the new genera- tion of laser printers. that enhanced the capability of VisiCalc. Lotus claimed that there were more than a thousand products available from 450 different suppliers. enabled data to be entered into a spreadsheet by means of a standard form. THE R I S E AND R I S E OF THE S P R E A D S H E E T 337 The market for complementary products probably began with VisiPlot and VisiTrend. it was a separate program with all the diffi- culties that communicating between two programs involved. An add-in was different to an add-on. The add-in market boomed. which saved the user from having to use scissors and glue when printing large spreadsheets.

which are worksheet models developed to solve specific types of problems.338 THE R I S E AND RISE OF THE S P R E A D S H E E T Now something truly remarkable and unexpected occurred. We have identified some 200 Lotus-related template. the user could zoom into the spreadsheet with different levels of magnification. Almost unwittingly. These can be loaded into 1-2-3. TaxAide. By mid-1985 there were more than a dozen products in just this category. or by taking over the producer directly. and there was help to detect errors. rather like a technological Just so story. going under names such as 1-2-3-Tax. And that. one could annotate cells. it was time consuming and error prone. is how the spreadsheet got its modern form. one could create dialogs for data entry. however. but application templates. TaxCalc. The most numerous complementary products for spreadsheets. An article in Software news in spring 1985 explained: The software enhancements to Lotus products come in many flavors.. Lotus 1-2-3 now incorporated advanced desktop publishing capabilities. and then 'filled in' with data. Lotus 1-2-3 had been transformed from. $50 for a ready made template was thus money well spent. and Lotus responded to these strong market signals by acquiring the rights to the best- selling add-ins. add-on. In the early 1990s. A small num- ber of add-ins (Allways among them) proved extremely popular.17 A typical and very popular type of application template was for tax prepa- ration. and these con- tinue to have a market niche. a stand-alone software product into a technological system. add-in and standalone soft- ware products.. While most users could in principle have created a spreadsheet for their tax return. which enabled it to maintain a 70 per cent market share despite intense competition from Excel and Quattro Pro. and the wince-inducing TaxPertise. By far the largest number of products are the many 'pre-solved problems' or templates as they are called. by which the market determined the appropriate mix of capabilities through its choice of add-ins. In the late 1980s Lotus started to sell 1-2-3 bundled with Allways and other popular add-ins. The Software news article cited above gave a taxonomy of four template . In the 1990s the phenomenon of add-ins vanished as quickly as it had risen. All of these features had once been provided by add-ins. were not add-ons or add-ins. the features of the most popular add-ins were seamlessly integrated into 1-2-3 and the original source of these capabilities was no longer evident.

real estate.18 Intellectual property and plagiarism The code of a program. he has 'made a difference'. Hence Dan Bricklin never became fabulously wealthy. The publishers of VisiCalc retained a patent attorney in 1979. real estate. existing software product by "reverse engineering'. with each sub-divided into different branches: General Business: financial analysis. but until recent times patent law has explicitly excluded algorithms and mathematical concepts. for many years the Craftsman Book Company of America supplied books such as The practical lumber computer. very few software patents had been granted. To take just one example. When Dan Bricklin invented VisiCalc. tax planning and preparation Utility Templates: multi-products sets. THE R I S E AND R I S E OF THE S P R E A D S H E E T 339 categories. oil production. VisiCalc was subsequently protected only by copyright and trademark law. These books consisted largely of mathematical tables. engineering management Vertical Markets: television and film. aviation. sales tools. sales & marketing. Estimating tables for home building. banking If one were looking for evidence of the extent to which the spreadsheet has replaced the table. and the advice given was that applying for a patent would be expensive and unlikely to succeed. business forecasting Personal Organization and Finance: personnel investment. accounting.The Craftsman Book Company now publishes just one compendium of tables. but several spreadsheet templates. asset management. although he has earned a place in history as the inventor of a device that created an industry and changed the working lives of millions of individuals—in his own words.19 It was relatively easy to clone an. office management. Today the first two of the these titles have vanished. This could be done at a much lower cost than that of . insurance.—whether expressed in binary or symbolic form— has always been protected by ordinary copyright law. while the role of the third is filled by Estimating with Microsoft Excel—unlock- ing the powerfor home Imilders. project management. and The national construction estimator. The practical rafter calcu- lator. and those that were usually relied on hardware-related processes. cash management. production management. it is in the realm of template publishing. budget.

but user research. In court Paperback Software testified that it had planned to enter the spreadsheet market in 1983.20 Another prominent 1-2-3 clone. failed product launches. Paperback Software subsequently ceased trading. but decided before launching the product in 1984 that it would need to be folly compatible with Lotus \ -2-3 in order to compete with it. and the cost of market building. This decision was of great importance to Lotus (and all other software makers) because for the first time it gave legal protection to an important Fig. These spreadsheet programs included menys compatible with Lotus 1-2-3 so that users could immediately make use of the software without learning new commands. The products were withdrawn from the market following legal action by the Lotus Development Corporation.340 THE R I S E AND RISE OF THE SPREADSHEET developing the existing program.) . (Copyright Eleanor Robson. the Lotus Development Corporation pursued a landmark look-and-feel lawsuit against Paperback Software International.The Twin. was quickly pulled from the market by its publisher Mosaic Software. Important protection against such cloning was provided by 'look-and-feel' case law. 12. because the development costs of the original consisted not just of code writing. Lotus was awarded costs and damages and VP-Planner had to be withdrawn from the market. the judge ruled in June 1990 thatVP-Planner's 'keystroke for keystroke' copying of 1-2-3s interface infringed Lotus's intellectual property rights.6 Lotus 1-2-3 clones. In 1987. Look-and-feel protection had been granted in a handful of cases dating from the 1950s for products such as greetings cards and TV quiz-show formats. After protracted legal argument. This firm had created a spreadsheet similar to Lotus 1-2-3 called VP-Planner. which it sold for $99—a fifth of the price of 1-2-3.

In one well known case. Errors in spreadsheets When using mathematical tables there were two important sources of error: incorrect values in a table. Lotus Development hurriedly turned out 1-2-3 Release 2.21 . Lotus sued Borland. A few days after the Lotus—Paperback Software decision. Quattro *s case was somewhat differ- ent.) Lotus pursued the case to the Supreme Court which con- firmed the decision of the Court of Appeals in 1996. where economic and safety considerations have long transcended any design whims of manufacturers. the Court of Appeals accepting that the protection of menus would impede user learn- ing and program interoperability. but the decision was reversed on appeal a few months later. THE R I S E AND R I S E OF THE S P R E A D S H E E T 341 aspect of user lock-in. Release 2. (This public spirited result was analogous to the use of a universal control layout in motor cars. Quattro had its own distinctive user interface. however. This was provided so that users of 1-2-3 could painlessly switch to Quattro without learning an entirely new set of commands. launched in September 1985.01 in July 1986. This did not produce an incorrect mathematical result (which would have been a major PR disaster) but failed to compute anything at all. One reason for this was that the cost of a product recall was very high so that spreadsheet publishers engaged in extensive internal and external testing before releasing a product. familiarity with a product's user interface.01 was offered to Release 2 users as a $15 upgrade (to cover postage and handling). the publishers of the Quattro spreadsheet. There are few if any recorded instances of mathematical errors produced by bugs in a spreadsheet program. The situation with spreadsheets is analogous. while errors by users were minimized by discip- lined calculating regimens. Nonetheless: Since [these] functions were at the heart of many a spreadsheet analysis. but it also allowed the user optionally to select a 1-2-3 style menu and command structure. Borland initially lost the case. which was generally considered to be better than that of 1-2-3. had an error in some of its financial functions. and mistakes made in hand computation. The current position appears to be that the look-and-feel of a program is protected only insofar as this does not stifle innovation by impeding a user's reasonable desire to switch to a competing product. Incorrect values in tables were generally eliminated by publishing errata sheets and new editions. Lotus 1-2-3 Release 2.

In both cases there is plainly a rhetorical element because if programs and spreadsheets were really so bad. but how many' errors they contained. and even fewer use them. The Auditor. above) enabled individual cells to be annota- ted so that the developer could create an audit trail as the worksheet was created. and users stumbled along doing their uninformed best. but had become major business applications in their own right.342 THE R I S E AND RISE OF THE SPREADSHEET User spreadsheet error was a much more serious problem. For example. Most academics. it would not be possible to use them. This revealed that 21 per cent of spreadsheets had serious errors that caused them to produce incorrect results. There was at first no easy way to avoid spreadsheet errors. However. to verify the relationships between cells in a graphical form. and that it was really a question of'not if. and conscientious users had to evolve a culture of building in check sums to ensure the validity of a spread- sheet. The literature on spreadsheet errors seems to share the characteristics of the gloomy literature on programming errors. will have encountered spreadsheet errors in the tabulation of student grades. These products. including The Analyst. The first major study was conducted by Price Waterhouse in 1987. Another set of add-ins (such as Noteworthy. the problem of reliability has been partly solved by the substitution of reliable software . which all worked in much the same way. During the late 1980s a number of empirical investigations of user- created spreadsheets were undertaken. and Cell-Mate. The modern spreadsheet now has auditing facilities fully integrated. while The Auditor was bought by the publisher of SuperCalc. the second issue of Lotus magazine devoted its cover story to the topic. and so on. While one presumes that most of the errors eventually come to light and are fixed.22 As the add-in market developed in the mid-1980s. one can never be wholly sure that a large-scale spreadsheet is error free. As one writer noted. then and now. nor does one feel confident in rearranging rows or columns without carefully checking a sample of results afterwards. to unravel circular definitions.23 Such reports prompted the manufacturers of spreadsheets to pursue the add-in makers to acquire their expertise. The company that made The Analyst was acquired by Lotus. for example. but they are so buried beneath a morass of features that few users know about them. enabled users to see behind the spreadsheet. a number of products came onto the market for error analysis. there was no obligation to check for errors. The problem of spreadsheet error was evident very early on. by the late 1980s spreadsheets had gone far beyond an infor- mal scratchpad. In the case of programs.

He wrote a cheque on the spot. there has been a debate about the so-called 'product- ivity paradox*—the fact that despite huge investments in information tech- nology there appears to be no concomitant. According to the journalist Robert Slater: Allen Sneider. upgrading programs. The more he used the program. keeping the hardware working. making disk copies. There is considerable anecdotal evidence thatVisiCalc was first used by middle-managers for preparing budgets and business plans. and Sneider knew that this was the tool he had been looking for. He asked senior management to introduce more Apples and VisiCalc. As Thomas Landauer has observed in his polemical book The trouble with computers: Spreadsheet programs can tempt users into endless puttering—changing this. We know comparatively little about their use. For many years. The time it took to run applications on a time-sharing system was thereby cut by eighty per cent. Whether the time spent using spreadsheets was economically useful is moot.24 Dan Bricklin has recalled that he was regularly buttonholed at computer shows by users who claimed that they were now able to do in minutes what had formerly taken hours. learning and teaching how to do new operations. organising and cleaning up overloaded file systems. THE R I S E AND R I S E OF THE S P R E A D S H E E T 343 products written under controlled conditions for end-user programming. Additional time is stolen by housekeeping: filing results. finding disks and files later. Users of spreadsheets As with tables. and community colleges offer courses in their use. A friend showed him a test copy of VisiCalc at the local computer store. the more impressed he became.25 . putting the com- puters and programs to work in different departments. An analogous role for spreadsheets has been played by application templates. The spreadsheet revolution was on its way. At the micro level there is a anecdotal evidence of a misap- plication of human effort to optimize resources at the margin. most historical writing has focused on the production and technology of the spreadsheet. was the first commercial user of VisiCalc. trying that—of little value. measurable increase in GNP at the macro level. on the grounds that two per cent of anything big enough is worth saving. Bookstores stock fat volumes about spreadsheet programs. He bought an Apple computer in 1978 and was trying without much success to run financial models on it. a partner at the accounting firm of Laventhol and Horwath in Boston.

For example the most frequent correspondents were enthusiasts explaining a 'macro' for some obscure process such as turning decimal numbers into Roman numerals. and most populous. The best by far was Lotus magazine for 1-2-3 users. Excel in 21 days. etc. there is the ephemeral peri- odical literature that surrounded spreadsheets. A second genre consisted of books designed for particular professions. These typically had titles such as Numerical analysis with spread- sheets. doctors. genre consisted of general purpose texts for using a proprietary spreadsheet—these typically had tides such as Lotus 1-2-3 for idiots. and sometimes tailored to a particular product. The first. Before that there was Spreadsheet magazine forVisiCalc users. and there are at least two untapped information sources to mine. with groups in every major US city. The correspondence columns of these magazines tell us much about spreadsheet users and their world. agriculture. These typically had titles such as Management accounting in Lotus 1-2-3. it quickly became the analytical tool of choice for numerate managers everywhere. economics. especially the man- agement and accounting professions but also in the sciences. Probably the best proxy we have is the textbook literature. there were user groups. and most other numerate professions one could think of. The third genre was college textbooks that used a spreadsheet as a pedagogical device for learning a par- ticular subject. Spreadsheet applications for analytical chemists. Financial accounting—-a spreadsheet approach. etc. geology. and after there was Excel user. First. These were not people one would want to spend a long train journey with. and soon many other professions. but other letters came from actuaries. Lotus 1-2-3 was the best served. and .344 THE R I S E AND RISE OF THE SPREADSHEET It seems that the spreadsheet initially began as an ad hoc tool. published monthly from May 1985 for nearly a decade. Among the 1000-plus spreadsheet texts in the Library of Congress catalogue one can discern three broad categories that had begun to emerge by the mid-1980s. There is a lack of statistical data on the diffusion of the spreadsheet in the various professions and education. Although a deeper analysis of spreadsheet users is beyond the scope of this chapter. By the mid-1980s college and university courses had also begun to use the spreadsheet as a classroom tool for teaching numerical methods. with different editions for each of the popular spreadsheets. and quantitative business school courses of all kinds. there is a huge potential for research. engineers. engineering. etc. Books were sometimes written for a 'generic* spreadsheet. However. economics. the sciences. rather like the electronic calculator of the early 1970s. Second. and many other specialities. engineering.

1670—4. The intended readership is usuaEy explicitly stated on a book's cover. Beginners* books usually have a title that hints at the readership level. the intermediate or advanced user. Further reading An excellent overview of current spreadsheet technology and practice in given by Deane Arganbright's 'Spreadsheet' entry in the Encyclopedia of com- puter science (4th edition. Texts for seasoned users are much longer—700 or 800 pages being quite usual—in order to cover the myriad features of the software. London. Fire in the valley: the mak- ing of the personal compiler. Reilly. D. One of the earliest and best accounts is given in Paul Frieberger and Michael Swaine. At user-group meetings a local spreadsheet whiz might explain. for example. McGraw-Hill. Calif. New York. 2000. how he (usually a he) had analysed sales data for his employer. Knepper. The silent majority simply used the tools at hand—whether logarithms or spreadsheets—and got on with the job. These are usually long and expensive. 1991.) There is vast publishing industry associated with propriety spreadsheets. Finally. The best account of the history of Microsoft's Excel is given in Daniel Ichbiah and Susan L. A. and add-ins. THE R I S E AND R I S E OF THE S P R E A D S H E E T 345 international chapters in Europe and Japan. 1984 (revised edition 1999). Texts are aimed at three levels of user: the beginner. There was never a user group for logarithm tables. particularly the market leader Microsoft Excel. and the professional developer. E. Hemmendinger. pp. ed. or is implicit in the length of the book and the prose style. At all times the writers of letters and attendees of meetings were a minority. but that perhaps says more about the times in which we live than about comparative tech- nologies. there are professional texts for developers of spreadsheet applications. though not always as annoying as Excel for dummies. but the company's magazine Lotus gives a superb coverage of the spreadsheet in its heyday. and D. and a peep between the covers will be enough to convince most readers that this level of expertise is far beyond what they could ever need in everyday spreadsheet use. The early history of the spreadsheet andVisiCalc is a staple of every popu- lar history of the personal computer. There is no good history of the Lotus 1-2-3 spreadsheet. Nature Publishing. Ralston. Rocklin.) . Books are typically 200—300 pages in length. templates.. The making of Microsoft (Prima Publishing.

February 1986. Anon. Mass. but the best starting point is one of the books on writing reliable spreadsheets such as Ronny Richardson. Landauer. Englewood Cliffs. A refreshing and sceptical discussion of spreadsheets and computers generally is given in Thomas K. Cambridge. Sigel and the staff of Communications Trends. 67-8. A good account of the Lotus Development Corporation's look-and- feel lawsuits is given in Lawrence D. 7.. 2nd edn. 4.' Management accounting. T. Margolis. 38.J. Conn. Lotus. Clarendon Press. Knowledge Industry Publications.W. esp. 92—9. E. A history of the software industry. Early scientific computing in Britain. 2003. 2. Washington. pp. E Bell. Basic cost accounting.. Oakland. Fire in the valley: the making of the personal computer. J. 1984-6. p. L. Best. February 1990. March 1973. 1984. p. Krueger and ]. 8. market dynamics in personal computer software. New York. D. Information for decision making. pp. 172-81.. Cognitive psychology. The software revolution. SPA worldunde data program for the year 1996. White Plains. Calif. Minneapolis. P. A good discussion of increasing-returns economics is given in B.346 THE R I S E AND RISE OF THE S P R E A D S H E E T A competitive analysis of the spreadsheet industry appears in Stanley J. McGraw-Hill. 9.138. DC. New York. McGraw-Hill. Greenwich. 'Positive feedbacks in the economy'. Mass.Arthur. 10. pp. 5. 1999). Winners. Software Publishers Association. 1990. Kaplan.. players. Grant and L. Liebowitz and Stephen E. Cambridge. New York. 1996). The trouble with computers (MIT Press. Campbell-Kelly. R. A. North-Holland. 1964. 1 I . M. E. Frieberger and M.~We$t Publishing Co. . 1985. Swaine. 1996). S. p. 1984. Prentice Hall. 178. Legal battles that shaped the computer industry (Quorum Books. M. Fertig.'Some future trends in spreadsheets'. 1999). BBL Basic Business Language: report generation and financial modeling. trends. 229-30. Rappaport (ed). Ctoaiken.. 1972-—reprinted in A. avail- able on the TYMCOM-X system. 1995. Notes 1. 103. Scientific American. There is some periodical literature on errors in spreadsheets. 12. NJ. M. Kohlmeier.. Cupertino. G. Calif. MIT Press. 6. Business /professional microcomputer software market. 1982. 3.. Professional's guide to robust spreadsheets (Manning. 95.Tymshare Inc. Conn. Oxford. 172. 'Financial Modeling and "What If" Accounting. B.. p.. From airlines reservations to Sonic the Hedgehog. NY. losers and Microsoft: competition and antitrust in high technology (The Independent Institute. Graham. Westport.

L. Quorum Books. Levering. Conatser... 40. Katz. New American Library. 24. Portraits in silicon. Moskowitz. 18. L. Ibid. Software news. J. O'Reilly. computer entrepreneurs: who's making it big and how in America's upstart industry. Microsoft secrets. pp. D. Selby. 1995. New York. 291—2. T H E R I S E A N D R I S E O F T H E S P R E A D S H E E T3 4 7 13. . 25.59—62. R. Excel 97 annoyances.craftsoian-book. pp.Wash. Programmers at work: interviews with 19 programmers who shaped the computer industry. Mass. 1996. The Free Press. Legal battles that shaped the computer industry. 38-45. R. and M.5. Lotus. Graham. 23. Slater. K. June 1992. Cusumano and R. Landauer.W Carroll. New York. MIT Press. 56-68. pp. 285—94. 22. Interviews with Dan Bricklin appear in: R. Lee. usability. A. 16. The. Redmond. and productivity. pp. 17. 1987. Information taken from the website of the Craftsman Book Company. 1984.July 1. 150.. M. 15-21. Hudspeth andT. 1986. Slater. 246. Westport. pp. 128—33. R. 1999. M. passim.. 130-51. Journal of end user computing 10:2 (1988). p. 1. The trouble with computers: usefulness. Microsoft secrets. 19.J.. K. 1997.'Avoid these common spreadsheet errors'. Conn.corn—accessed July 2001. Cambridge. Susan Lammers. 20.'1-2-3 through the years'. 'What we know about spreadsheet errors'.'Lotus enhancements blossom forth'. L. March 1985. Lotus. www. 21. Mass. esp. Calif. MIT Press. D. 14. p. T. 66-71. W. R. Cusumano and Selby. Tempus- Microsoft Press. Sebastapol. Grushcow. Panko.985. Woody. Portraits in silicon. Cambridge.

This page intentionally left blank .

and Fellow of Kellogg College. Mary Croarken is on the Editorial Board of the IEEE Annals of the History of Computing and is Honorary Secretary of the British Society for the History of Mathematics. Let Newton Bel (1998). Biographical notes Martin Campbell-Kelly is reader in computer science at the University of Warwick. co-authored with William Aspray. He was formerly editor of the journals Annals of Science and History and Philosophy of Logic. (MIT Press. set theories and the foundations of mathematics from Cantor through Russell to Godel (Princeton University Press. He is editor of the Works ofBabbage (Pickering & Chatto. where he specializes in the history of computing. due in 2004. 1989). Mobius and his band (1993) and Oxford figures (2000) published by Oxford University Press. 1870-1940: logics. Her publications have centred on the history of com- puting and mathematical tables in the pre-1950 period and include Early scientific computing in Britain. 2003). Ivor Grattan-Guinness is Professor of the History of Mathematics and Logic at Middlesex University. (Clarendon Press.D. He is an Associate Editor (mathematicians and statisticians) for the New dictionary of national biography. Oxford University. He is a co-editor of The nature of time (Blackwells. She is currently a Visiting Fellow at Warwick University. Raymond Flood is University Lecturer in Computing Studies and Mathematics at the Department for Continuing Education. Mary Croarken gained a Ph. 1996). His books include the two-volume Companion encyclopedia of the history and philosophy of the math- ematical sciences (Routledge. His main research interests lie in statistics and history of mathematics. 1989). 1990). 1994) and The search for mathematical roots. and ICL: a business and technical his- tory (Oxford University Press. 1986). His publica- tions include Computer: a history of the information machine (Basic Books. 2000). He has recently completed Fmm airlines reservations to Sonic the Hedgehog: a history of the software industry. He is editing for Elsevier a large collection . in History of Computing from the University of Warwick in 1986.

An actuary by train- ing. a research and archival center. Norberg holds the ERA Land-Grant Chair in History of Technology in the University of Minnesota. he is an authority on risk management and he received the Finlaison medal of the Institute of Actuaries in 1999. His latest book. and the Royal Society'. He worked for many years as an archivist at the Public Record Office in London. where he is also Professor of Computer Science. He is Editor-in-chief elect of the IEEE Annals of the History of Computing and has published articles in that journal as well as Chance. Mr Lewin has contributed sev- eral papers on actuarial science to BSHM conferences. 193-208. He is the author. Notes and Records of the Royal Society 51 (1995). case studies 1640—1940. O'Neill. which will be published by Princeton University Press in 2003. He is the author of 'Joseph Moxon. Chris Lewin is Head of UK pensions at Unilever pic. FRS. 1996) and of a number of articles in the history of astronomy and in the history of computing. Graham Jagger is an associate lecturer with the Open University. His book Pensions and insurance before 1800-—a social history will be published byTuckwell Press in 2003. Arthur L. The information state in England: state gathering of information on citizens since 1500. His publications include numerous articles on the early history of actuarial science and insurance in The Actuary And other actuarial magazines. with Judy E. 1962—1986 (Johns Hopkins University Press. He is also Director of the Charles Babbage Institute for the History of Information Processing. . the Communications of the ACM and Inference. where he is carrying out research into activities of the seventeenth century British mathematical community. He was editor of History and electronic artefacts (OUP 1998).350 B I O G R A P H I C A L NOTES of essays on Landmark writings in Western mathematics. David Grier is an associate professor of computer science and international affairs at the George Washington University. also to appear in 2004. of Transforming computing technology: informa- tion processing jar the Pentagon. the American Mathematics Monthly. He is currently writing a book on human computers. is currently in the press. Edward Higgs is a senior lecturer in history at the University of Essex. His publications include Making sense of the census (HMSO 1989).

He is an authority on calculating machinery. and languages at the Oriental Institute. Brown. He is a former editor-in-chief of the Annals of the History of Computing. Books he has authored include The dream machine: exploring the computer age (with Jon Palfreman. Williams is professor emeritus of the Department of Computer Science. University of Calgary and Head Curator at the Computer History Museum in Mountain View California. 2000. He was active in the International Astronomical Union and served as pres- ident of Commissions 4 (ephemerides) and 5 (documentation and astro- nomical data). an historian of technology. Michael R.Viking Penguin. 2002). Until recently he was Assistant Director & Head of Collections at the Science Museum. history. He is currently an Honorary University Fellow in the School of Mathematics at Exeter and participates in the activities of the Norman Lockyer Observatory at Sidmouth. Doron Swade is an engineer. Oxford. He was Head of the Almanacs and Time Division of the Royal Greenwich Observatory from 1974 to 1989. and now works as a professional clarinettist. 1. George Wilkins joined H M Nautical Almanac Office in 1951 and served as Superintendent from 1970 to 1989. His most recent book is The cogwheel brain: Charles Babbage and the quest to build the first computer (Little. Margaret de Valois gained a BSc in Pure Mathematics at the University of Warwick. London. He has published widely on curatorship and the history of computing. specializing in the letters of Sir Isaac Newton. BBC Books.999). 2100-1600 BC: technical constants in bureaucracy and education (Clarendon Press. and his many publications include A history of computing technology (1987 and 1995). . Oxford. She works on the intellectual history of ancient Iraq and teaches ancient Near Eastern archaeology. where she studied the history of mathematics with David Fowler. he was also chairman of working groups on numerical data and on the determination of the rotation of the Earth. and a leading authority on the life and work of Charles Babbage. BIOGRAPHICAL NOTES 351 Eleanor Robson is a Fellow of All Souls College. She is the author of Mesopotamian mathematics. 1991). Ms de Valois has written for various industry publications and The Actuary magazine.

This page intentionally left blank .

37 Babbage.295-316 Burroughs 302 Astronomical cuneiform texts 37—41 Ellis. Index Abramowitz.281 . 36-41. Howard 138.94 Bauschinger. 28.20. George Biddell 157.11-12.222 al-Khwarizmi 6 Babbage.8.193.$ temps Bachelder.295-316 Airy. Arthur. 272 Ballistics Research Laboratory 11 Analytical engine 12. P. 270. Adelard of Bath 6 181. 111-112.193. Charles 8-9. 193. Arthur 202 Akkad 22 Akkadian language 22. 93-94. king of Assyria 29. 139 Almanacs Babylon 23. 37.282 Apple II 330 Bennett.34 Bessel.254 Astronomical tables 7. H. 133. Ashurbanipa!.192. 149-151. 189. Aljamim tables 6 157-163. god of Uruk 37 Bell Laboratories 124.239-240. 328 Annuities 87.202 Auwer.41 see ephemerides Babylonia 13.36. Cambridge University BASIC 326 126 Basic Business Language (BBL) 326.247. 300.John 133 see Nautical almanac Bagay. Aubrey. 69.278.245 Society Airey.286 Berlin Academy 185 Aramaic language 27.25. 52. 202 Assurance 80. 136. 198 Admiralty 193. FriedrichW.287 Bessel function tables 12.35. al-Battani 6 123. Julius 135—136 Anu.132.187 Astronomers.26. Daniel 96 Archibald. Valentin 118 American ephemeris Baily. 256. 243.151 Ballistics tables 283 Analytical Society. 139.287 Applied Mathematics Panel 282. 35.192-193. 177-203. Arnott.60.98-99 planetary tables 39-40 Actuarial tables 79-99 procedure texts (precepts) 39—40 Adab 29 Astronomical observations 178—203 AdawiSjJohn Couch 193 comparison with tables 181.245. 191.98 Titlmkic Rfgiarnvntanae 192 Assyria 20.296 177-203. 106. 5-6.196. Francis 8 American Mathematical Society 269.126-129. 89. John 60 197. 302. 191-192. 305 Astronomical Research Institute.29 Bernoulli. Accounting tables 20—27 lunar tables 38-40 Actuarial profession 5. Paris 107. Berlin 135 Admiralty Computing Service 9.297. 184.39-41.241.156-157.286 Ashur 26.303 Astronomical Society.A. US Baily.311 ephemerides 39. see Royal Astronomical Aiken.90.Jobn 241-242.197.244. 40.245.110. 34.79.27.27. 258.129. Milton 283-284 Astronomer Royal 8.35 198. Hans 266.28. 28. 89.295 117. 10.15.41 302-303. see National diaries 38 National 137-138. 8 see Nautical dlnuitwtc. 8. 41. 183. Enuma Anu Ellil 37-38. 37 see Coniun&ance 4e. Academic des Sciences.42.14. Raymond Clare 132. 29. 38 Bethc. 283. Neil 221 244.198. A. Accounting machines 244.39.34-35.

Blundeville. 112 Calculators.39.9. 203.10. .64—65 of 1861 209 Tfigonometrica Britannica 49—50.297.66. Alexis 189. Arthur 236-238. human Leyton Arithmometer 300 Computers.354 INDEX Bickley. 188 Colebrooke. Computation 3.247. 188. 296. John Cayley.192. M. Burroughs Adding Machine 300—302 328-330. 323-345 Nova Brunsviga 244 Computers.87. Charles IW.179. 272-289 Cambridge University 52. 98.227.190.33. Lazure 108 Bridgewater.25-26.301. 145. 341 Comptometer 131 see also computers. 343 Carnoc. 94. 299. Brunsviga 136-137. 196.297.308 50-52. 108 Clement.190. Frank 247 Bureau de Cadastre 10. 224-225 118.179.244. Thomas 56 250. Brunsviga-Dopla 137. 239. I I I . Tycho 7. 283. 111. 279 Centesimal system 111.191. Mercedes Euklid 135 225.202 Carlini.198 Trinity College 66 Bradley. England and Wales 4.225 British Association for the Advancement ot Census. 12.201.108-110.194.145-170.10.259.192. 69. 315.344 Biot.2.193. Dan 327-328. Janet 50 Observatory 193 Bourcfaier. 229 l^ogimlhmvmm chilias prima 50.33. king of England 49. 5.105. Earl oft see Egerton.316.219.11. Joseph 127-8. 58—59. Johann T.70 Members 246 Checking tables 109. 339. 202.Johann K. British Association Mathematical Tables Philadelphia. 15.146.195. 151-153. Henry 58 St John's College 54 Bouvard. 150-159.302 189.180. 257-260. electronic 5. 296.279.145-147. 197 Cleaver. 305-307. Lyman 270-271.247. 197. 213. 280. 250 Bothwell. 52 Carlisle table 93 Bricklin. Arithmetics fagtiritkmifa 57.137. 201. Edwin 221 Chairmen 241 Charles II.296. 56-58.C.Vannevar 269 Columbia University 271.323. James 189.188. William 248-249.310. 117 see computers. 65-66. (1876) 132-134 Committee 8. 64-72 Census.11.278.257 Borda.315. US 12 Science 8. 69 of 1901 213 Briggs. Arithmometer 146-147.12.203.70. 69 209-214. human Blanche.284 of 1911 12. 68.284.24. 54-62. 254-255 Sunstrand 281 Bilk of Mortality (Carlisle) 89 see also accounting machines Bills of Mortality (London) 80-86 see also punched card machines Binding 145. Secretaries 241 304 Brookhaven National Laboratory 286-287 Church.281 302. William 147 Burckhardt. human Blagden.1. electronic 50.235-249 Centennial International Exhibition. 197 CODATA 314 Burg.180. 188 Calculators. Gertrude 266-267.96-97.32.326. 253 Briggs. Odhner 137 150-152. electronic Edmondson 240 see also computers. Compound interest tables 80—83 225. 247. 98. Monroe 280. Coast and Geodetic Survey 268 189. 11.117 Bessel Function Sub-Committee 245 Chadwick. 300. human 1. 252. 235-261. 270. 224-225. Francesco 197 Biahe. 54.117-118 King's College 71 Borland (company) 341 Mathematical Laboratory 248.20.107.126. 302.195.31.221-222. 1. Henry Thomas 8 Bush.147 Calculation see Computation Binet.B. 139 Bureau des Longitudes 107-109. 65—67.279 Compound interest 79—80 Calculating machines i 1.244 215. Henry 7. Jacques P.J.J. 49-50. 62.225.111-112.240.

108.M. 136-137 Comrie. 297. 196-197.296 Hamann's 135—136 sec also hairdressers Minor.194. 59. Arthur 241 Cuneiform script 22. 189. . 69. 197 De Colmar.265-266 203.0639 21. 123-140.153-157. James 89 Cugerus 32 Doodson. 110. 37 De Prony. 159.10. 137-138. 138. Humphrey 127 Encke. 163-170. Thomas Rowe 91—92 Dale.97.160. see also Nautical almanac.308-310 Connaksance ties temps 7. 243-247.151. Edward 2H4 302. 254 Ecole des Pouts et Chaussees 106.7. 41 Downing. king of Assyria 29 Babbage's 1833 demonstration piece 128 Euler. Gaspard Ricfae 4.222-224 De Moivre.240.181.27.150-152. Wallace 279 VAT 12593 27.238. 15.117 159-160. 1 159.251. 192. Augustus 99.150. 160.250. (DEC) 327 see also ephemerides Disease classification systems 216-219 Continuous Mortality Investigation Bureau 94 Distribution media for tables 314-315 Cowell.194 Babbage's Difference Engine No. 185. Babbage's Difference Engine No. 240.244 Davis.112. 186.W. Abraham 87 ENIAC 288 De Morgan.Johann f.247. 29. 300 AUAM 73.EH. 222-224 255.159 307-310.0400 23-24 Dunkin.146. 66. Antoine H8 Errors in tables 14.296.184. Palin 93 Darwin.2 129. Charles 198 Equitable Life Assurance Society 89.184. Depot Generalc de la Guerre 111.284.325 Differences 11.153-156 Enlil.194 Edmonds.111.244-245. 3000-2400 ucti) 22. 279. Allan 240-242. Scheutz' 129-131. A. Harold 269 Ernden Functions 243 Davy.315.195-198.252 Elliptic Functions 238. 108 Cunningham. 277.163 188.190.John. 299.307. William 90 Egerton. James 286 113-116. 87. INDEX 355 224-225.William 300 AUAM 73. 67. John 277. 23. Leonhard 115-116.222-224. 222.Jean Baptiste Joseph 111.179-180. W.190. Ephernerides 5. 341-343.114.108-110. first Earl of Bridgewater 60-61.297 Esarhaddon. 149.228.30-31 Ecole de Geographes 107. 72. 18-42 Dudley Observatory. 203 Digital Equipment Corp.68.62 Dartmouth College 132-133 Elderton.238.?!> Gonntus&inct' de$ temps Delambre. Excel user magazine 344 139.148-149.107.23-24 BM 34083 40 Early Dynastic period (r. Condon. 53. Charles 249.183. Edwin 297. 72. Grant's 132-135 273-289.248.105-118.300-303.304.Albany.192 see also astronomical cuneiform texts Delaunay. 112 Curtis.224 English life tables 131. French.139. 325. 127. Thomas 146.346 159.118. 287 Ecole Polytechnique 106. god of Nippur 19.29-30 CBS 3323 18-19 East India Company 52 Plimpton 322 33. 93 Deparcieux.304.124-125. Conant. Wiberg 131-132. 300 Distribution of tables 147 Craftsman Book Company 339 Dodson. and US Death certificate 227 196-197 Decimalization 3 see <?/. Leslie John 12.265-266. Design of tables 307-310 274-277. New York 130-131 A 681 29 Dunkin. 117-118.163. 224. 285. Alfred 136 comparison of British.310-312.295-316 Deacon. Difference engines 11-12.90-92. Department oi Oriental Pyrenees 107 117.163-170 Exponential tables 238 . 300 Cuneiform tablets 14.34 Eckert. CBS 2124 32 27.

Andrew 49 First World War 269 Harvard Mark I 12. Strasbourg 135 236-240. Eroma 10 IBM-compatible PC 331. Dorr.26.91. 55-56.150. Christel 12. king of Babylon 23 Finlaisonjohn 89-90.245 Harvard University 12. US 282 and Stationery Office 224 Gibbs. C. Edmund 7. 282. 198.224 Hydrographic Office.225—226 Search Room 211 >w tifej punched card machines Statistical m>ieu> 216.138. Micaiah 241 General Board of Health 222 Hind.253 Imperial College.66. 310 Gresham College 54-56.13.62 Jet Propulsion Laboratory 306. 61 Jones.197. 247 Heyshamjohn 89 Galton. 131 Halley. E. 90.38 International Astronomical Union 243.John 8. Sir Thomas 54-55 see Mesopotamia Groornhridge. 180. Samuel 287 Hammurabi. 124. 50. 126-127.314 GreenhiU.356 INDEX Factor tables 239.33 Guard digits 151 Gunter. 327 Parlies magazine 337 Heliocentric position co-ordinates 179.244. 224 Institpt de France '107 Grant.F.244. 258 . Richard C. Benjamin 8.208-230 Hipparchus 13 Annual report 216.334 Goodwin.118 Felt. 68. 156. 300 Census report 210 Hollerith Dictionary of occupational terms 208-209 Census Machine 12.312-313 Gififord. 281. James 238-239. goddess of Untk 22 Gompertz.91—92 Increasing returns theory 331—332. 72 Iraq Gresharn.John 300 General Registrar Office 12.227 Hitchins. Princeton 271 Greeks 29.229 Hairdressers 109.255-258 Indiana University 269. C. Stephen 8 kin 23. Bob 327 184 French Revolution 185.135-136 Finkelstein. 60.W. William 139.T. London 305 Glaisber. London (1851) 224 Institute of Advanced Study. 212.186 Henderson. Frankston. 220. Alfred 239 Interpolation 113.Wolcott 133 IBM Corp.118. George. 308. Malachy 299.68 Hill. Ronald 241.240 Hahn.William 239 Gamier J. see Royal Greenwich Observatory Gallon Eugenics Laboratory.96 Filon. 156 Hill. T. UCL 228. 300-301 and Treasury 213-215. Edmund 8. Philipp MattfaSus 124 Fair. George Bernard 12.G.114-115. 60-65. 59.60—65 England) 50 Dt* s&tore et radio 60. John 243 Friendly Societies 94-95 Herschel. James Whitbread Lee 155. 194-195. George William 201 GelUbrand.245 Fisher. 333 Glaisher. 108 Highland Society of Scotland 91. Fippard.133.138." 252. Imperial Observatory.95 Gauss. Peter Andreas 193. 61-62. king of Scotland (fames I. Charles "E. Henry 55-56.132-135 Institute for Numerical Analysis 287 Graunt. 66 James VI. 224.281 Goulbura. 303-306. 93 199-200 FirminDidot 111-112 Hart. king of Canon triangulomw 50. 228 Hudson.157-158 Funk Software (company) 336—337 Herstmonceux Castle. Louis 241 Halley's Comet 39 Financial analysis software 326—328 Hairann.66.314 Gunter's Sector 60.John 83—85 Institute of Actuaries 98—99 Great Exhibition. Francis 227 Hicks. C.95 Hansen. 253 Inana. Henry 157 Initial values 274 Graham. 245. 86-87.

107.66. 34 57-58.196. Urbain J.266-267. 250 metrological lists and tables 33-35.62-64 Kersseboom. 65. R. 127. god of Babylon 23. 193.301.48-72. 52—53.345 202.72 Kaplan. Miles 197 Three-point interpolation formula 72 Look-and-feel 340-341.146. 191.315 98131 Massachusetts Institute ofTechnology (MIT) Linear programming 288 258.William 62. problems with Napier's definition 54. 71-72 Kassite period (c. 117 Makeham. 191.42 .29-34.189. 11. Thomas 214 coefficient lists 33. Gottfried Wilheim von 10. Henry W. . 203 Lowen.66.299.Wyman. 42 Lodge.297.299.69-72 Lotus 1-2-3 331-335. 152. Marduk.69-70 Laderoianjaek 284 popularization of 56. Patrick 8 of trigonometric functions to base 10 50. 37 199-200.126 division of the quadrant 66 Konigsberg 202 of natural numbers to base ten 50.156.68. 65.201. 29.271-289 M'canique celeste 188.34 sec also astronomical cuneiform texts Law of mortality 91-93. William 91 Legendre.189. tables of 188. 296. 69-72 Lagrange.111. oi trigonometric functions 49.193.344. 70 Marshak. 197.67. Manchester University 247.70 Klingenberg.282 Linear zigzag function 38. 14. Library of Alexandria 6 Maskelyne. 108. Eleanor 240 39-40. Pierre Alexandra Francisque.71 Kelly. 41 Mathematical cuneiform texts Lister.83 105-118. Arnold 8.300. 158 198 Larsa 23. 39-40.116 Manchester Royal Infirmary 228 Legendrian Functions 238 Manchester Unity Friendly Society 95 Leibniz.80. Life assurance tables 5.jenldn 93 Logarithms 5.wr navigation tables Language of tables Los Alamos 283.67. 28-29.194-195. Joseph Louis 106. Poul 80 oi trigonometric functions with decimal Klipstein. and W. 250 150 Manly. Lodge.199. 59. Jones.69.199 Lunar motion. 198. Latin 60.346 for unified set) 188. Dionysius 129-130.337-338. 1600-1100 BOB) 18-19.M.203 Mari 33 Leybourn.345 Kalhu (Ninirud) 26.346 Lalande»Jerorne 7 LORAN. 90.185-189.71—72 Method of variation ot arbitrary constants London Mathematical Society 244 184. 280.182. 112—113. William 88 of trigonometric functions with decimal Kish 35 division of the degree 59.10. 97 Leverrier.63. Lardner.33. 295-296. 93-94.200 Longstretb.184-189.63 Kanesh 34 differences 53. Robert 281. Philipp Engel 124.98. 65. 35. 56—57 195 rules for use 63.67. 69.61. JerroM 328 examples of use 52.157.241-242. 52.323.194. 179 59.7.151. 334-335.336. 328.332.189.J.135-136. Nevil 7.236.70.34 Liverpool University 247. Lotus magazine 342.269.A.202.25.198.98 see also astronomical tables Lee.10. 199 Lotus Development Corp. Keplerjohannes 7.37 artificial numbers 52.253 multiplication lists and tables 27. 157. Laplacian programme (see also precepts: search 340-341.124.72 344-345 Laplsce. 26. 95.66. Alfred 239-240.189.68.286 English 52.136.340-341. 97 see also ephemerides Lefort. Pierre Simon 182.57-59.83-91. INDEX 357 jones.

193. 25-26. G.196-198.37.274. 240-241. 11.225.72 Number-divisor tables 244 Constmctio (Mirifid lagarithmotum atnonis. 193. 203. John.195.11.257.279. 12. US 176-177. 10. 188 see al$o ephemerides Mechanization 123-140. 344. king of Babylonia 29 Microsoft Windows 332.247. 295-316 Mayer. William 227 .29. Nautiml almanac.35. British 7.286 Newtonian gravitation 187 Mortality tables 92-94 Newtonian mechanics 188 Mosaic Software (company) 340 Newton. Caspar 86 Monetary functions 96—98 New York University 267 Moore School of Electrical Engineering 11 Newcomb. 266. Tobias 7. Phillip 280.203 Microsoft Corp. 52. De&riptio (Mirifid logaritknt&twn canonis descriptio) 219-221 48-49.32-3. 334. Milan 197 34-35. 115-116 Nirnrud: see Kaihu Moxon.284-285. 338.249. god of Nippur 25 Multiple decrement tables 95—97 Nippur 18-19.29.243-245.330-335. 540-320 257-258 BCE) 27.288 Mathematical symbolism 70 National Cash Register (company) 137 Matkemalifal ttihles and other aids to computation National Mercantile Lite Assurance Society 93 138. eighth Lord Merchiston 7.38 Miller.247-249.93 Napier. 305. 255.146 Nimirta.310. Lord: $ee Napier. comtruttio) 52 Occupational classification systems 216. 200. 333. 258. tabular calculations 33 278.178-179 Morse.37 Miilkr.152.256 National Physical Laboratory 248. Nautical almanac. 303. Committee 270.301-313 200. Mesopotamia (ancient Iracj) 10.18—42 60. Isaac 7.201. Committee 197.203 Menabrea. Mathematical Tables and Other Aids to 256. 192. Number theory tables 239.35.Joseph 70—72 Nineveh 27. 335 Neo-Assyrian period (c.40 National Bureau of Standards 257.145.315 Mathematical Tables Project. New York 8. British 9. 23. John Navigation and navigation tables 5. Jeffrey 242-244. 244 48-54. 335 LORAN 266. Luigi 151 see also ephemerides Merchiston. 345 Nebuchadnezzar II. 177.Johann Helfrich 124. see British Association Mathematical Tables 137-138.286 Rabdologiae 50-51 Ogle. 156-157.252. 265-289 189.61-63.252.188. 24.270.189. 52. 29—30. Malcolm 267—270 Newton. seventh Lord Merchiston 50 Nosology 216-219 Napier.196-198.56.158-170. 56-58. Joshua 89 Neumann.198.11.282-283 Microsoft Excel 9.61-62.300 Non-polynomial functions 150 Northampton life table 88-89.195. see accounting tables of squares and square roots 27.56-58.John 69-70 Mouton. 900-610 BCn) 27.281.201. 282.1.69 Occupational dictionary 208—209 Napier's bones 50 Office for Scientific Research and Development Promptuary 51 282-284. 250. Archibald. 279. 313 Computation. 198-203 Morrow. machines 33.13—14. 59. Robert 52 reciprocal lists and tables 31—355 40 National Academy of Sciences 267-270.26. 287 Natuif 252 Mathematical Tables Committee Nautical Almanac OfHce.203. Simon 180.) Napier.287 tables of inverse cubes 33 National Accounting Machine.65.37 Milne.65. 29. Nee-Babylonian and Persian periods (r. 9.35.358 INDEX Mathematical cuneiform texts (contd.41 Multiplication tables 236.196.295-316 Committee Superintendents 301 see Royal Society Mathematical Tables Nautical Almanac Office. 224-225. US 176-177.

189.183.188. 198. tables of 180. Benjamin 133.203 Probabilities of survival 80 Oughtred. 222 Synfaxis 6 Paris Qbservatoire 112-113.163 Royal Greenwich Observatory 189.310-311 Protector Assurance Society 9 Paper 148. JohannTheodor 135—136 Registration and Marriage Acts (1836) 211 Philip II. 202 135 short-term usefulness of 181.198. 185-188.163 Ptolemy.126. theory of 191 Depositary of Unpublished Tables 252 Present value 81—82 General Sub-Committee 251 Price Waterhouse (company) 342 Mathematical Tables Series 252 Price. Bartholomew 239 Royal Swedish Academy of Sciences 130 Price. Pascal. Poulkova Observatory 198 300.Wilhdm 191 Prieur de la Cote d'Or. Berlin equations of 178-181.247 corrections to 184. 240.39 Flong 148 Almagest 6. Robert 97 178.195 Royal Society 55.202 Public Health Act (1848) 222 Parseval. 13.247.53. 34 Pell. 249-260 search for unified set 178. 296. 279. 279.194. 198. Royal Society Mathematical Tables Committee 195. 197 Roosevelt.196.184. Orbital elements 180-181. perturbation theory Recorde. 2000-600 BCE) Printing of tables 124.129-130. 55 Polynomial functions 150.180. Claude 106 Plana. 189. John 66. Karl 9. Giovanni A.132-134. William. 198. Arthur 313 Canon dottrhwe triatigulamm 53 Piazzi. 198.202 249-260 sec al$« Laplace: Laplacian programme Bessel Function Panel 258 Precession. 297. 25-26. 146.179. Rothamsted Agricultural Experimental Station 200.31-34.238.191.303. 185. Franklin 268. 147. 192.202 Opm Palatinum 10.184. 154-158.191.300 sec also astronomical cuneiform texts Royal College of Physicians.182-183.A. A.332 Ready reckoners 1.John 97 Pivotal values 150. 202 Registrar General 131 Peters. Giuseppe 191.196. 86. 23-24 Pearson. Richard 88-89.152 Riche de Prony. C. 6061.198. 193.306 Powers. 192 Royal Astronomical Calculating Institute.190.64. 147-149.195. 64.198 Royal Astronomical Society 8. Parthians 29. 188.36. INDEX 359 Olbers.194-195.200.311.227-228 Pythagoras' theorem 33.A.70-72 Probability function tables 274 Trigonometric (Trigonifinelry) 70—71 Proofreading 145.190.127.296. 199-200. 108 Old Babylonian period (c. 108 Punched card machines 12. 222.186. Lord 239 Personal Software (company) 327. 145. 302-304.179.66 Pierce.202.282.297.181.134 Radio Shack TRS-80 330 Pension funds 95—97 Rayfeigh. Stockholm 130 .181. Claudius 6.191.37 281.158. 282. 185.72 Pennsylvania University 11.37 Precepts 179. Simeon 188 Royal Exchange 54.M. 289 Planetary motion. 191.197. king of Spain 50 Rheticus 10. B.297.165-170.311-313 192. London 214 Poisson. tables of 244 Royal patronage 35.202.66 Philips.199 Rhodes. 146 313.314 Pattern wheels 161 Puzrish-Dagan 21. 224. Ida 284 Pitisens 65 Richards.95 Royal Technological Institute. Regiomontanus 64 193.193.201.299 193.197.41 136-139. 201 244.3—4 Perturbations ol the planets.192.161. Pondjohn 156.39 Paperback Software (company) 340-341 Handy tables 6 Paris Exposition (1855) 130.

342 138.192. 300. Microsoft Excel Second World War 247. Thomas Southwood 221 Sumerian language 21—22.29-30 Stevin. 338. 36.V.160-161. Quattro Pro 335.152.27. Alexander 137.133.308-310 Smith. 15. 139 Sideways 336-337 Scientific Computing Service Ltd.332 TJ. 30. Samsu-iluna.342 Schicfcard. 36 Smithsonian Institution.29-35. 131 Susa 33 Snyder. 197 SuperCalc 342 see ab« astronomical tables The Twin 340 Seleucid period (r. Nicolaas 88 Smith. Smith. London The Analyst 337. 252.341-343. London.334 Secular variations. 219 Shuruppag 27. king of Babylon 23 335-339 Sang.337 Scribes 14. C. Donald 243-244. Henry 236 Sumer 20-22 Smith.239 errors 325. Irene 284 Shadivppum 33 Stereotyping 148-149.129-131.36 VisiTrend 334. H. 322-345 Thomas. Watson Computing Laboratory 279 Software news 338—339 Tabling sheets 213 Solar motion.25-29.284 Spreadsheet (complementary products) 334. 337 132.245 The Auditor 342 Scribal schooling 20. Bebr Georg and Edvard 12. king of Assyria 29.297. 22. 14.. Edward 69. 337 Bel-ban-apB 35 Spreadsheet (products). see also VisiCalc. 60 Cell-Mate 342 Scheutz. 296. Simon 81 Sickness tables 94-95 Stocks.154. 320-120 BCE) 39-42 VP-Planner 340 Sennacherib.Virgil 272 Syllabaries and lexical lists 35—36 Society of Apothecaries.11 ? Add-ons and add-ins 337 Sargon II. Ernest C.37 Allways 337-338 Sargon. F>.35.23. London 214 Software Arts (company) 327. Taxonomy ot tables 2—5 203 Taylor's theorem 116 see aha ephemerides Temples 18-20.360 INDEX Sadler. WilEam 300 Smart. Sir Henry 55.225 . 302. Thomas 87 Stokes. Johannes Nikolaus 91 Spreadsheet 2. William (Lord Kelvin) 236. 108 Subtabulation 150.John 87 Struyck.224 Noteworthy 337.A.Washington.279.163.W. 128.24.252 clones 339-341 Thomson. Adam 10. 42 VisiPlot 334. Percy 228 Simpson. E 245 165-170 Shulgi. 346 Three-body problem 179 etymology 325-326 Ticking method 213. David Eugene 9 304. D.310 user groups 344—345 Salzcr. tables of 189.W 146 SeeMore 337 Science Museum.35 Stokes.33. Richard 71-72 Slide rule 225 Stratford. Lotus 1- Nammah 31 2-3.303.T. investigation ot 1H5—187. 275-277.37 Speidelljohn 69 Tetens.E. king oi Assyria 29 S^ffdrfSfe'f magazine 344 Service Geographique de 1'Armce 117 Standard Tables Program 94 Sexagesimal place value system 31.111 Stegnn.190-191. king of Akkad 22 application template 338 Savile.282-283.41.314 Microsoft Multiplan 330.197.21.249. 97 anatomy of 329—330 Thompson. macros 344 303.222. 5.255.113-114.341 195. Herbert 265-266. George 236 Sippar 26. Sheppard.C.136. king of Ur 22 Stevenson.

Wilkins.15.105-118.66.22. 334. Maurice 250. 66. 306. 52-54.71-72. Joseph 197 34.23-24.305.31. 69 Writing boards 29.147-149.163. Winlock. Franz X. University College.153-154.278 US Naval Observatory 298. Wargenrin.61. Sears 198 Trigonometric functions and tables 10. John 280. New "York Ur 21. INDEX 361 Toddjohn 257 Vlacq. 149-151 Ttenchantjean 80-82 Walker.38. 59. 189 Transcription 145-146.244.82-83.41 Van Orstrand. 288 Tontine 80.E. 68. 285. Ward. Edmund 64 in cuneiform 18-19. Martin 12. 36-37.198 Tonn.W.41 Witt. Lorenzo 80 von Neumann.311 Wingate. 147.97 tabular lists (cuneiform) 25. 65.89 von Zach.14.37.327-328. Pehr 88 236.325. 284 Yale University Observatory 305 VisiCalc 322-323.131-132 Typesetting 145. Richard 79. Archbishop James ofArmargh 56. 189. 27. Adrian 65—66 "IHediin tables 6 von Lindenau. 20.282 Worksheets 275-277.32-33.36. 64 63-64.53.22.21. Edward 52.179. 282-283 WPA see Works Project Administration Ussher.37.285 337. Alfred 95 definition of 53 Weaver. William Thomas 60 307-308. Yeshiva University 271. 256.B. Wittstein. 56.33.258-260 Typography of tables 2. Thomas 91-92.49. 66.70 Wiberg. George Udny 227-228 . Warren 286 logarithms of: see logarithms What-if analysis 326 spherical trigonometry 61—64. 151.26. 299-300 VisiCorp (company) 332 Yule.35 Worksheet (alternative term for spreadsheet) Uruk 20-21. 62-64. 339 Young.307. 286. C.269-270. George 310 61.35. 70. John.300 Watson. 312. Theodor 96 38-39 Woolhouse.58 Wright. 252. 299.166 Wilkes. 91 Works Project Administration (WPA) 8. 285 Veblen. Bernhard A.S.23-24. World Health Organisation 228 314 World Wide Web 335 US Navy 266.160. London 137 267-289 &Y also Galton Eugenics Laboratory see alw Mathematical Tables Project. Oswald 9.39 325-326 US Army 266.330-332. 40.33.281.11.67.